Home / New releases / October 2026

⚡ New AI tool releases: October 2026

All 107 verified AI tool, model and product launches from October 2026, newest first and grouped by week. Each entry links to the official announcement; dates come from the vendor's post. Last updated 2026-10-30.

Jump to week: Oct 26–Nov 1, 2026 · Oct 19–25, 2026 · Oct 12–18, 2026

Week of Oct 26–Nov 1, 2026

Full week page (29 launches) →

2026-10-30

Drex Max v1 open weights (Nace AI)

Nace AI open-sourced Drex Max v1, a dense 27B multimodal decision model. Instead of generating text, it reads a context (text, JSON or images) plus one or more typed questions and returns a probability for every option, up to 255, in a single forward pass, which suits jobs like ticket routing, yes/no checks and ratings. Nace reports 67.52 on the public Decision Index 0.4 and 64 on the Artificial Analysis Decision Index. Weights are on Hugging Face under the Nace.AI Open RAIL-M license in bf16, Q8_0 GGUF and affine Q4 formats; the Q4 build runs locally on Apple silicon Macs with 24 GB of memory (MLX or Nace's Kapitsa), and hosted inference is also available.

Source ↗
2026-10-30

Mage-VL-Cowork open weights (Microsoft)

Microsoft's Mage team released Mage-VL-Cowork, a 4B-parameter computer-use agent model that decides at each step whether to click through the screen (screenshot-grounded mouse and keyboard actions) or to write code and terminal commands, inside one policy. Microsoft reports 66.0% on OSWorld-Verified with a 50-step budget, ahead of much larger open agents such as Qwen3.6-27B (63.9%). The weights are on Hugging Face (microsoft/Mage-VL-Cowork) under Apache-2.0, with a technical report and training recipe on GitHub.

Source ↗
2026-10-30

mischief 1 image-to-video model (The Biological Computing Co.)

The Biological Computing Co. (TBC) launched mischief 1, an image-to-video model built on the open Wan 2.2 base and tuned with efficiencies TBC says it measured in living neurons grown in its lab. TBC says it matches the base model's quality while generating about 5x faster. It costs $0.005 per second of video (2.5 cents for a five-second clip), which TBC says makes it the lowest-priced image-to-video model on both fal and OpenRouter, where it is live now (480p/720p, 3- or 5-second clips from a first-frame image).

Source ↗
2026-10-29

Mistral Large 4 (Le Chonk) open weights

Mistral released the weights of an improved Mistral Large 4 ("Le Chonk"), its 1-trillion-parameter, natively multimodal mixture-of-experts model, which entered public preview earlier in October. After three more weeks of reinforcement learning, Mistral says it leads open-weight models on Cognition's FrontierCode 2.0 (19.7%) and Terminal-Bench 4.0 (41%), scores 48% on the Artificial Analysis Coding Agent Index (tied with GLM-5.3 while using 47% fewer tokens) and 51% on its Cyber Index, and now offers low, medium and high reasoning modes. Weights are on Hugging Face under the Mistral Ecosystem Open-Weight license (a modified MIT license; Model-as-a-Service businesses need a separate commercial license). The API costs $0.70 per million input tokens and $2.10 per million output tokens, and the model is also on OpenRouter, Microsoft Foundry and NVIDIA NIM.

Source ↗
2026-10-29

Pika H3 Flex open weights (Pika)

Pika released H3 Flex, its post-trained version of MiniMax's H3 video model that generates in six steps instead of 50. Pika says it can make a 15-second clip with synchronized audio at 768p in as little as seven seconds of transformer compute, works from up to 12 references, exports up to five separate audio stems (dialogue, score, SFX, Foley, ambience) and finishes at up to 4K. The open release (generator adapters and inference code) covers text-to-video, first/last-frame and reference-to-video at 480p and 768p; the 1080p–4K upsampler and stem separation stay Pika-hosted. It is available now through Pika's API and comes to the Pika creation app on November 2. Pika says it ranked #1 among open-weight models (#7 overall) on DesignArena's text-to-video board at launch.

Source ↗
2026-10-29

Canva Brush (Canva)

Canva launched Brush, its first vision-language model for design, built by Canva Research a week after its Canva Craft image model. Brush reads a design's layout and layers and makes the edits you ask for, from a single line of copy to every slide in a deck, and everything it produces stays fully editable. Canva says it is up to 69% faster on general design edits and 46% cheaper per edit than its previous design agent. It powers supported edits in Canva AI for Canva Pro subscribers starting today (pick "Brush" in the Canva AI dropdown), with wider availability and Brand Kit support coming.

Source ↗
2026-10-29

SFX v3 (ElevenLabs)

ElevenLabs released SFX v3, its newest text-to-sound-effects model. It follows prompts that describe several sounds more reliably (for example a crowd cheering, a ball dropping and the crowd booing), handles music and vocals better, and produces fewer artifacts on sounds like gongs, chimes and whistles than SFX v2, according to ElevenLabs. It is available in ElevenCreative (Sound Effects tab, Flows, Studio and the Music Editor), in the ElevenLabs mobile app and in the API as eleven_text_to_sound_v3, where the docs mark it as limited access that may not be available in every region.

Source ↗
2026-10-29

Krea Ad Studio (Krea)

Krea released Ad Studio, the first "Krea 3 Expert" (a version of Krea 3 tuned for one job). You paste your website and it builds an editable brand kit (logo, colors, fonts, tone of voice, selling points) and a product catalog, including Shopify catalogs, then generates up to 12 static ads per batch for Instagram feed (4:5, 1:1), Stories and Reels (9:16) and landscape (16:9). You can change any ad by asking Krea Agent, and Ad Studio also works through the Krea MCP server in Claude. It is live on every paid Krea plan from Basic up; brand research and product import are free, and ads use compute units.

Source ↗
2026-10-28

Prism speech-to-speech model (Rime)

Rime launched Prism, a speech-to-speech model for voice agents that writes an inspectable text version of each reply before it speaks, so guardrails can check every word while the audio keeps cues like hesitation or frustration. Rime says it answers in about 300 ms, runs tool calls in parallel with speech, scores 56.6 on Artificial Analysis's τ-Voice harness (ahead of non-reasoning realtime models from OpenAI and Google) and can be fine-tuned, with self-serve RL and SFT APIs being built with beta customers. It can run in Rime's cloud, a private cloud or on-premises. Rime's Free plan includes $50 of usage a month across its models; on the pay-as-you-go Flex plan Prism costs $0.080 per minute.

Source ↗
2026-10-28

GitHub Copilot Fusion GA

GitHub made Copilot Fusion generally available at GitHub Universe. Instead of picking a model, you pick Fusion at Standard, High or Max effort and it routes each prompt to the right models, sometimes chaining them (a cheap draft escalated to a stronger model, or a draft reviewed by a second model). GitHub says Fusion Max scored 60 on the Artificial Analysis Coding Agent Index, within two points of Claude Code with Opus 5.5 Max, at about half the API cost per task. It works on all Copilot plans in VS Code, the Copilot app, Copilot CLI and JetBrains IDEs, billed per underlying model, with 30% off model costs for a limited time.

Source ↗
2026-10-28

Nemotron 3.5 Transcribe (NVIDIA)

NVIDIA released Nemotron 3.5 Transcribe, an open-weight, promptable speech model (about 34B total, 4.3B active, hybrid Mamba-Transformer MoE) that does transcription in 40 languages, speaker attribution, translation and audio Q&A in one checkpoint, on recordings up to an hour long. You choose the transcript style with a plain-language instruction. Weights are on Hugging Face and ModelScope (BF16, FP8, NVFP4 and GGUF) under the OpenMDW-1.1 license, with vLLM and SGLang support from day one, hosted endpoints on Baseten and DeepInfra, and the open Granary v2 data pipeline for fine-tuning.

Source ↗
2026-10-28

AssemblyAI-TTS-1

AssemblyAI launched AssemblyAI-TTS-1, its first text-to-speech model, aimed at voice agents. Audio starts streaming in about 90 ms, it offers 4,000+ voices and instant voice cloning, and it costs $15 per 1M characters with unlimited concurrency and no annual contract. It is available over HTTP, WebSocket and an OpenAI-compatible speech endpoint, and on OpenRouter.

Source ↗
2026-10-28

GitHub Agentic Workflows GA

GitHub Agentic Workflows, an open-source GitHub Next project, is now generally available. You describe an automation such as issue triage, CI failure analysis, docs updates or test writing in a plain Markdown file, and it compiles to standard GitHub Actions YAML that runs a coding agent (Copilot or your own) on your existing runners. Agents run read-only in a sandboxed container behind a firewall by default, and a threat-detection job checks proposed changes before they apply. Available on GitHub Free, Team and Enterprise; the default Copilot CLI engine needs a Copilot license.

Source ↗
2026-10-28

Replica open-source AI prototyping tool (GitHub)

GitHub Design open-sourced Replica (MIT license), an AI prototyping tool that turns natural-language ideas into interactive prototypes built from your team's own design-system components, patterns and docs. Designers set the guidance, and anyone on the team can create and refine prototypes from the GitHub Copilot app. Prototypes can be shared for comments on an optional, self-hosted Replica Hub.

Source ↗
2026-10-28

Particle-1 (Particle)

New lab Particle launched Particle-1, a software-engineering model post-trained from DeepSeek V4.1 Flash with reinforcement learning to use fewer tokens. Particle reports 28–52% fewer output tokens and 34–52% lower cost per task than the base model on Terminal-Bench 2.1, ReactBench, ADE Bench and DeepSWE 1.1, with pass rates within the base model's confidence intervals. It is available today through the Particle API at $0.30 per 1M input tokens and $1.20 per 1M output tokens.

Source ↗
2026-10-28

Mistral Responses API (public preview)

Mistral introduced a Responses API in public preview for building agents. Each response returns the model's answer, its reasoning and its tool calls as separate items, which makes multi-step agent loops easier to handle than with Chat Completions. It follows the OpenAI Responses format and works with all chat models hosted by Mistral, including Mistral Large 4, with streaming, JSON Schema output, client-side function calling and image/file input.

Source ↗
2026-10-28

Clover-35B-A3B-Preview open weights (DiffuSpace)

DiffuSpace (founded by members of HKU NLP's Dream team) released Clover-35B-A3B-Preview, an open-weight diffusion language model that generates text in parallel blocks instead of one token at a time. It is a 35B-total, 3B-active MoE with a thinking mode and 128K context, reported at 680 tokens/s on a single GB300 GPU (BF16) and 73.6 on SWE-bench Verified. Weights are Apache-2.0 on Hugging Face and ModelScope, with SGLang and llama.cpp support.

Source ↗
2026-10-27

LFM3 (Liquid AI)

Liquid AI launched LFM3, a family of six open-weight ~2B models built on one new LFM3-1.7B-Base backbone, covering text, vision, audio, omni and two decision models. The text model fits in about 1 GB at 4-bit and has a 256K context window. Every model can be fine-tuned on a single GPU. Available on Hugging Face, the Liquid Playground and the Liquid AI API, with llama.cpp, MLX and vLLM support (and on Ollama the same day).

Source ↗
2026-10-27

Echo-3 + SpAItial app (SpAItial)

SpAItial released Echo-3, a world model that turns a few minutes of iPhone capture into a photorealistic, metric-scale 3D Gaussian-splat world, filling in areas the camera missed. You can measure inside the world and share it by link to any browser. It ships with a new SpAItial iOS app (Android coming soon) on subscription plans, with the first month free. Enterprise API access is offered; processing costs 300 credits per captured minute (900-credit minimum).

Source ↗
2026-10-27

Ling-3.1-flash-1019 open weights (InclusionAI / Ant Group)

Ant Group's Ling team open-sourced Ling-3.1-flash-1019 under the MIT license. It is a post-trained upgrade of Ling-3.1-flash (560B total / 25B active hybrid-linear MoE) that scores 44 on the Artificial Analysis Intelligence Index, up from 41, with gains in agentic and scientific tasks. The model card claims a 1M-token context. Also released: Ling-3.1-flash-base, a BF16 base checkpoint for further training. The FP8 weights are about 574 GB.

Source ↗
2026-10-27

Runway Ads GA

Runway made Runway Ads generally available on every paid plan. It connects your Meta, Google and TikTok ad accounts, reads which ads perform best, and generates on-brand variants (localized and resized) from a working ad or a template using your Brand Kit. Every ad needs team approval before it publishes to the connected accounts. Enterprise admins enable it per workspace.

Source ↗
2026-10-27

MAI-Image-2.7-Preview (Microsoft AI)

Microsoft AI released MAI-Image-2.7-Preview, an image editing model for restyling or relighting scenes, editing text and symbols in place, and restoring or enhancing assets. It debuted at No. 2 on Artificial Analysis's image-editing leaderboard and No. 3 on Arena Edit. It is available to try free in MAI Playground; text-to-image is coming later, and there is no API or published price yet.

Source ↗
2026-10-27

Giotto Alto and Giotto Solo (Giotto.ai)

Swiss lab Giotto.ai (Lausanne) launched a family of reasoning models pitched as European sovereign AI: Giotto Alto, a flagship that runs on eight B300 GPUs, and Giotto Solo, which runs on a single GPU. Alto is free on Giotto's European shared cloud, including a no-account anonymous chat and new iOS and Android apps. API access is a flat fee; dedicated Swiss/EU cloud and on-premises licenses are also offered.

Source ↗
2026-10-27

v0 artifacts (Vercel)

v0 can now create artifacts: reports, charts, diagrams, datasets and interactive HTML pages that it builds and edits in the chat, kept separate from your app's code. Interactive explainers and charts appear inline, and Markdown reports and JSON/CSV data open in a tab next to Preview. HTML runs sandboxed, files can be up to 5 MB, and artifacts can be published to share.

Source ↗
2026-10-27

Delta for Claude Code: multiplayer sessions (Zed)

Zed's Delta made Claude Code multiplayer in beta. Run /delta share to get a link, and a teammate who runs /delta join enters the same thread from their own Claude Code session, with your conversation and uncommitted worktree changes synced across machines. It works through the Delta CLI and an open-source Claude Code plugin; more harnesses are planned, starting with Pi.

Source ↗
2026-10-27

Unlimited OCR 2 (Baidu)

Baidu open-sourced Unlimited OCR 2 (3B, MIT license), a document-parsing model whose Sparse Reference Sliding Window Attention keeps memory and per-step latency nearly constant, so it can parse documents of unlimited length, even a full book in one request. Baidu reports state-of-the-art results on multi-page and hundred-page benchmarks; runs with Transformers or SGLang, with a Hugging Face demo.

Source ↗
2026-10-26

GPT-6.1 Terra (OpenAI)

OpenAI released GPT-6.1 Terra, a fast, efficient model for visual reasoning, computer use, coding and natural writing, with standard API prices 50% below GPT-5.6 Terra ($1/$5 per 1M input/output tokens) and a 1.05M-token context window. Available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, in the API as gpt-6.1-terra, and the same day in GitHub Copilot and Amazon Bedrock.

Source ↗
2026-10-26

Runway App Studio

Runway launched App Studio: describe a creative app in plain language and it codes, tests and deploys it, chaining Runway's image, video and audio models with your Brand Kits. Apps can be kept private, shared with a team or published by link; it is included in every paid plan and draws on existing credits. The older Runway Apps now live in App Studio.

Source ↗
2026-10-26

Figma Weave models via the Figma MCP

Figma made every video and image model in Figma Weave (such as Seedance, Kling and Nano Banana) available to external AI agents through the Figma MCP server, so you can generate images and videos, including image-to-video, from your agent chat. Requires a Figma Weave account linked to Figma; generations use Weave credits.

Source ↗

Week of Oct 19–25, 2026

Full week page (45 launches) →

2026-10-25

Nanbeige4.3-33B (Nanbeige LLM Lab)

Nanbeige LLM Lab (BOSS Zhipin) open-sourced Nanbeige4.3-33B under Apache 2.0: a 33B dense model with a 512K-token context for agentic coding, tool use and office 'cowork' tasks, plus a Base checkpoint. Benchmarks are vendor-reported; a distilled Nanbeige4.3-3B is planned.

Source ↗
2026-10-25

Pomerado

YC Fall 2026 startup Pomerado launched an open-source service that turns any website, including legacy portals and internal tools, into a hosted API or MCP server for AI agents, built from a single prompt, with self-healing integrations as an alternative to browser agents. Free to try.

Source ↗
2026-10-23

Hy-Nomas interactive world model (Tencent Hunyuan)

Tencent Hunyuan open-sourced Hy-Nomas (Apache 2.0), an 8B interactive video world model that generates a continuous 720p, 24 FPS world one segment at a time while you steer it with new text instructions and camera paths. Includes weights, training and inference code, LoRA adapters (cinematic, robotics), a single-RTX-4090 INT8 setup, a web demo and the 280-case Hy-Nomas-Bench.

Source ↗
2026-10-23

Kimi K3.1 Pro + Highspeed (Moonshot AI)

Moonshot AI released Kimi K3.1 Pro, a 2.8T-parameter (104B active) multimodal MoE model with a 1M-token context, aimed at long-horizon coding, 3D modeling in Blender/CAD, financial research and agents, plus a 2–3× faster Highspeed variant. Live in Kimi apps, Kimi Work, Kimi Code and the Kimi API ($2/$10 per 1M input/output tokens); full open weights are promised by Nov 4.

Source ↗
2026-10-23

Nemotron 3.5 Bolt E4B 4GB (NVIDIA)

NVIDIA shipped Nemotron 3.5 Bolt E4B 4GB, the small open model announced Oct 15: post-trained from Gemma 4 E4B for local assistants, interactive characters and robots, fitting in about 4GB with multi-turn tool calling. Weights (4GB GGUF, BF16 and a DFlash draft model) are on Hugging Face.

Source ↗
2026-10-23

MiniMax M3.1 Flash open weights

MiniMax published the weights for M3.1 Flash (428B total, 23B active per token, 1M context, text/image/video input) on Hugging Face, four days after the Oct 19 API launch. Day-0 serving is on vLLM and Together AI. It is released under the MiniMax community license (attribution required; prior authorization above $20M annual revenue), not MIT/Apache.

Source ↗
2026-10-23

Hi3D V1.0 open source (formerly Sparc3D)

Hi3D open-sourced Hi3D V1.0 (formerly Sparc3D), an image-to-3D model trained on about 2M curated 3D assets with sparse-grid geometry reconstruction up to 1536³. Code is on GitHub and weights are on Hugging Face, with a free online demo. Hi3D also cut credit use on its hosted platform by over 50%.

Source ↗
2026-10-22

Canva Craft (Canva)

Canva released Craft, its first in-house image generation model from Canva Research. A two-stage design plans the composition and text before rendering images at up to 2K in 30+ aspect ratios, with text in 55 languages. Available to all Canva users in Dream Lab and via the Canva public API; Canva MCP support is coming.

Source ↗
2026-10-22

Ideogram Vector 1.0

Ideogram launched its first vector model: it generates clean, editable SVGs with smooth gradients and sharp typography from a text prompt, or converts an uploaded image into a scalable vector. Live in the Ideogram app, the API and partner platforms such as fal.

Source ↗
2026-10-22

Decisions API GA (OpenAI)

OpenAI made the Decisions API generally available on gpt-6-luna. POST /v1/decisions returns a probability, a choice or a score for bounded questions over text, images or JSON, up to 10x faster than the Responses API. GA adds prompt caching, native GPT-Live integration, US/EU inference residency and up to 400 questions per request.

Source ↗
2026-10-22

Framer MCP

Framer launched an official remote MCP server (mcp.framer.com/mcp, OAuth) so AI tools such as Claude, ChatGPT/Codex and Cursor can create pages, edit layouts and content, manage CMS collections and publish Framer sites. It is now the recommended way to connect external agents to Framer.

Source ↗
2026-10-22

Sume API (Sume Labs)

Sume Labs launched the Sume API, an image and video generation API (Seedance, Kling, Veo, Wan, Nano Banana and more) billed at the underlying provider's price with 0% markup. It is available via REST, a hosted MCP/CLI and a web studio, and the pricing page lists side-by-side rates against fal and Higgsfield.

Source ↗
2026-10-22

Ling-3.0-flash-Sante open weights (InclusionAI / Ant Group)

Ant Group's Ling team open-sourced Ling-3.0-flash-Sante, a healthcare-focused MoE model (124B total, 5.1B active per token) for medical reasoning, deep research and multi-agent workflows. It reports 54.29 on HealthBench Professional. Weights are on Hugging Face.

Source ↗
2026-10-22

Carbon-X (Hugging Face)

Hugging Face released Carbon-X, an open (Apache 2.0) 14B model mid-trained on 300B tokens of interleaved DNA and text, so you can ask about a sequence or mutation in plain English or generate DNA. It is for research, not clinical use, and runs locally.

Source ↗
2026-10-22

AgentSpot (ThoughtSpot)

ThoughtSpot launched AgentSpot, an agentic workforce platform where business users build, share and run AI agents and workflows grounded in governed company data ('multiplayer AI'), with outcome-based rather than token-based pricing.

Source ↗
2026-10-21

Claude Duet (Anthropic)

Anthropic opened a waitlisted research preview of Claude Duet, a proactive agent app (iOS, Android, claude.ai) powered by Claude Opus 5.5 that sends a morning brief, monitors connected apps, works on its own cloud computer and asks before sending, paying or deleting. Included in US Pro and Max plans.

Source ↗
2026-10-21

Flint (Recursive)

Recursive released Flint, an open-weight (Apache 2.0) 40B-parameter mixture-of-experts coding model with about 2.5B active parameters, built for fast local agentic coding (125+ tokens/s on an M4 Pro MacBook Pro) with day-0 support in vLLM, Ollama, llama.cpp and LM Studio.

Source ↗
2026-10-21

Mammoth-1 + Elephant-1 decision models (Base Labs / Baseten)

Base Labs released open-weight decision models mammoth-1 (27B) and elephant-1 (9B), post-trained from Qwen3.5, which return calibrated probabilities for routing, classification, yes/no and rubric questions in one pass. Mammoth-1 tops the Jev Decision Index; both are served on Baseten.

Source ↗
2026-10-21

OpenAI models on Google Cloud

Google Cloud added OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna to Vertex AI Model Garden in an invite-only private preview, usable via Google Cloud APIs and Codex workflows and payable with existing Google Cloud commitments. General availability is planned in the coming weeks.

Source ↗
2026-10-21

Breken

Breken launched a 'production before production' service that rebuilds your app in isolated worlds with synthetic users, aged data and failing dependencies, then reports reproducible bugs with traces and draft fixes. Its ESCAPES benchmark claims it caught 76 of 235 shipped bugs.

Source ↗
2026-10-21

Kanana-2 vision encoders (Kakao)

Kakao open-sourced two Kanana-2 vision encoders (426M 'medium' and 98M 'tiny') under the commercially usable Kanana Open License, claiming top results across 42 benchmarks, especially Korean OCR and document understanding, with 60%+ lower latency.

Source ↗
2026-10-21

Nemotron 3.5 Embed, Rerank + Parse (NVIDIA)

NVIDIA released Nemotron 3.5 Embed, Rerank and Parse at GTC Berlin: open multimodal retrieval models (embedding and reranking over text and page images, with NVFP4 variants) plus a document-parsing model, aimed at grounding enterprise agents in company data.

Source ↗
2026-10-20

Perplexity Computer, fully local on Mac

Perplexity's Computer agent can now run fully on-device in the Mac app, with the model (PPLX Pro or Qwen) and the agent working with local files and native apps like Notes, Mail and Numbers without using Computer credits. For Pro, Max and Enterprise on Apple silicon; 36GB+ memory recommended.

Source ↗
2026-10-20

EAGLE-4 speculative decoding

EAGLE-4 is an open-source speculative-decoding system with one trained drafter per target model that adapts draft-tree depth, width and verification budget to hardware and batch size. It reports about 4x faster decoding on Qwen3.8-27B, with SGLang and vLLM integrations for NVIDIA B300 and AMD MI355X.

Source ↗
2026-10-20

Voxtral 3 (Mistral)

Mistral launched Voxtral 3: new speech-to-text models (Voxtral Small/Mini Transcribe 3, incl. realtime), Voxtral Mini TTS 3 with emotion and pace control and voice cloning ($13 per 1M characters), and a voice agent SDK, via its EU-hosted API or self-deployment.

Source ↗
2026-10-20

Eleven Realtime (ElevenLabs)

ElevenLabs introduced Eleven Realtime, end-to-end speech-to-speech models for voice agents: v1 Mini for natural conversation and v1 Max for complex tool-using tasks, with 90+ languages, 17,000+ voices and an OpenAI-compatible Realtime API.

Source ↗
2026-10-20

Fibo 2.0 (Bria)

Bria released Fibo 2.0, an 8B open-weights image model trained only on licensed data that unifies generation and instruction-based editing (up to four reference images, masks, JSON controls); Dev and Edit Turbo weights are on Hugging Face and it ranks #1 among open-weights models on Artificial Analysis image editing.

Source ↗
2026-10-20

Ornith-1.5-122B

Ornith released Ornith-1.5-122B, an MIT-licensed open-weight MoE model for coding and agentic work that runs on a single NVIDIA DGX card, with FP8, NVFP4, GGUF and MLX variants and day-one support in Ollama, SGLang and Hermes Agent.

Source ↗
2026-10-20

Ampersand v0.1 (humans&)

humans& released its first model, Ampersand v0.1, a collaborator model trained in long multi-party simulations to help groups reach shared goals (and to stay quiet when it should); built from GLM 5.3 and available now via an OpenAI-compatible Agents API.

Source ↗
2026-10-20

Phonon on-device TTS (Gradium)

Gradium open-sourced Phonon, tiny on-device text-to-speech models (31M and 40M parameters, the smallest about 64 MB) that run offline on phones, laptops, browsers and embedded hardware; the English 31M model is Apache 2.0.

Source ↗
2026-10-20

FastH3 Omni (Hao AI Lab FastVideo)

UCSD's Hao AI Lab open-sourced FastH3 Omni, a distilled, accelerated version of MiniMax H3 reference-to-video that takes text plus image, video and audio references and generates synced video with audio in eight steps, with NVFP4 checkpoints for local RTX/Spark GPUs.

Source ↗
2026-10-20

Beacon (Asymptote Labs)

Asymptote Labs launched Beacon, an MIT-licensed, local-first memory and telemetry layer that captures sessions from 20+ coding agents (Claude Code, Cursor, Codex and more) and turns them into reusable skills, alongside a $5.5M seed round.

Source ↗
2026-10-20

Gemini tools for small businesses (Google)

Google added a small-business landing page to the Gemini app with on-brand visuals (upload a brand kit, paired with Nano Banana 2.1), financial insights from messy spreadsheets and receipts, and connections to Google Business Profile, Ads, Workspace and apps such as Canva, Shopify and Xero.

Source ↗
2026-10-20

VARCO Sound (NC AI)

NC AI officially released VARCO Sound, a multitrack generative AI sound platform that creates and edits scene-appropriate audio from text, images and video, with a plugin for existing production workflows.

Source ↗
2026-10-20

Mintlify Autopilot

Mintlify launched Autopilot in beta on Pro and Enterprise plans: an always-on docs agent that watches your repos plus Slack, project trackers, support tickets and feature flags, drafts doc updates when something changes and opens a pull request for review.

Source ↗
2026-10-20

llm-d 1.0

The CNCF-governed llm-d project (founded by Red Hat, IBM, Google Cloud, CoreWeave and NVIDIA) released llm-d 1.0, a hardened open-source core for distributed LLM inference aimed at production and context-heavy agentic workloads.

Source ↗
2026-10-19

MiniMax M3.1 Flash

MiniMax released M3.1 Flash (revealed as the anonymous 'Space Bunny' model): a 428B MoE with 1M-token context for agents, coding, computer use and security, ~3x faster than M3 at $0.15/$0.60 per 1M input/output tokens; on the MiniMax API, M Plan and MiniMax Code, weights due later.

Source ↗
2026-10-19

Mercury 3 (Inception)

Inception launched Mercury 3, its flagship diffusion LLM: 1,068 tokens/sec on NVIDIA GPUs, 1M-token context, tool calling and structured outputs; $0.20/$0.70 per 1M tokens (50% off at launch) on its API, OpenRouter, Baseten and Vercel.

Source ↗
2026-10-19

Clairic (Cloudflare Workers AI)

Cloudflare previewed Clairic on Workers AI (@cf/cloudflare/clairic), an extraction model that returns typed values (text spans, numbers, dates, estimates, forecasts) with confidence scores instead of generated text; $0.10 per 1M input tokens, open weights planned.

Source ↗
2026-10-19

Viggle JST-2

Viggle launched JST-2, a physics-native 3D world model that generates explorable scenes, animation-ready rigged characters from a single image and full-body motion from video in one shared 3D world; free to try and available via the Viggle API.

Source ↗
2026-10-19

FLUX 3 Max (fal)

fal launched FLUX 3 Max, a text/image/keyframe-to-video model post-trained with Black Forest Labs on FLUX 3; fal says it renders a 10-second clip in about 11 seconds, 22x faster than Seedance 2.5, at 720p and 1080p on fal's API.

Source ↗
2026-10-19

Exa 3.0 + Exa Lite

Exa released Exa 3.0, a major upgrade to its web search API for AI agents (larger embedding model, Dynamic Highlights GA using 95% fewer tokens, expanded index), plus a new Exa Lite search type at $4 per 1,000 requests.

Source ↗
2026-10-19

Together Post-Training

Together AI launched Together Post-Training, a managed service for custom SFT and reinforcement-learning loops on open-weight models (LoRA or full-weight, custom losses) on dedicated B200 GPUs, with a $9/GPU-hour launch promo.

Source ↗
2026-10-19

Maxwell (doubleAI)

doubleAI unveiled Maxwell, billed as the first 'code execution model': it runs inside live Python/PyTorch programs, swaps in faster hardware-optimized code and proves each replacement is semantically equivalent; early access now.

Source ↗
2026-10-19

Wabi 2.0

Replika founder Eugenia Kuyda's Wabi opened Wabi 2.0 to everyone on iOS, Android and Mac: a multiplayer AI messenger where you add friends or family to threads and the agent plans, books and builds mini-apps on demand.

Source ↗

Week of Oct 12–18, 2026

Full week page (33 launches) →

2026-10-18

endue OS for agents + secretary

endue moved its agents' cloud computers to endue OS, giving each agent its own Station with Linux and Chrome workspaces billed from one credit balance. It also added a per-account secretary that splits requests across agents, an Orbit view, and per-agent MCP access from Claude Code, Cursor and VS Code.

Source ↗
2026-10-17

DeepSeek-V4.1-Flash-1017 + Search & Fetch API

DeepSeek shipped a same-price update to V4.1 Flash with stronger agentic benchmarks and shorter outputs, live on web, app and API with MIT weights on Hugging Face. Its Search & Fetch APIs opened to everyone at $0.75 per 1,000 calls per endpoint.

Source ↗
2026-10-17

Aivah: Skills, Companion and Ask Aivah

Aivah relaunched its AI-employee workspace with Skills (reusable playbooks), meeting attendance in Meet, Teams, Zoom and Webex, a face-to-face Companion voice mode, Ask Aivah (Cmd/Ctrl+K) setup help, and one brand style for generated docs, decks, images and video.

Source ↗
2026-10-16

Beam (Reflection)

Reflection released Beam, its first open-weight model: a 501B-total / 23B-active sparse MoE built for coding, reasoning and long-horizon agents, with weights on Hugging Face under Apache 2.0 and access through Reflection's platform and 40+ launch partners.

Source ↗
2026-10-16

Glimpse (Elorian)

Elorian's first model, Glimpse, is a visual-thinking model that revisits images and video frames while it reasons (cropping, pulling frames, running code). It is available via API and chat UI at $0.60 / $3.00 per million input/output tokens.

Source ↗
2026-10-16

Ling-3.1-flash (InclusionAI / Ant Group)

Ant Group's InclusionAI published full materials and MIT-licensed fp8 weights for Ling-3.1-flash, a 560B-total / ~25B-active hybrid linear-attention MoE with a 1M-token context, aimed at coding and agent workflows, with day-0 SGLang and vLLM support.

Source ↗
2026-10-16

Perplexity Decider v1.2 (open-source decision models)

Perplexity open-sourced pplx-decider-flash-v1.2 (4B) and pplx-decider-v1.2 (27B), decision models that return probabilities instead of text, served through its Decisions API at about 1 and 2 cents per million tokens and also available on OpenRouter and Vercel AI Gateway.

Source ↗
2026-10-16

Pika Plugin for ChatGPT

Pika launched an official ChatGPT plugin that brings 40+ Pika creation tools (Video Studio, Camera Director, Motion Transfer, Product Ad and more) and 55+ image, video and audio models into ChatGPT's chat and side panel, with timeline editing and access to your Pika library.

Source ↗
2026-10-16

GPT-Live prompt-to-voice (OpenAI API)

OpenAI made prompt-to-voice generally available to all API customers: describe a voice in text to create a custom GPT-Live voice (free to create; speech billed at normal GPT-Live session rates). Nine new German, French, Hindi, Japanese and Korean voices were added.

Source ↗
2026-10-16

Replit Bring Your Own Cloud

Replit introduced Bring Your Own Cloud for AWS, Azure and Google Cloud: full BYOC (private preview) runs the Replit platform and published apps in the customer's own tenant, with BYOC for Published Apps and Image Exports (Docker, Terraform, SBOM) in early access.

Source ↗
2026-10-15

Kling 4.0

Kuaishou's Kling AI made Kling 4.0 generally available: 3–30 second videos at up to 4K with up to 10 keyframes and 15 references, in the refreshed kling.ai creator plus official MCP and CLI access; the developer API and 10-bit HDR are listed as coming soon.

Source ↗
2026-10-15

Muse Spark Open (Meta)

Meta released open weights for Muse Spark Open, a natively multimodal 816B-total / 42B-active MoE model with a 1M-token context window, built for coding and agentic work; weights on Hugging Face and hosted APIs via OpenRouter, Groq, Together AI and others.

Source ↗
2026-10-15

Krea 3

Krea released Krea 3, an image-editing model that returns most native-2K edits in under 8 seconds at $0.06 per 2K image ($0.03 at 1K), with multi-reference and regional edits; live on krea.ai and via fal, Replicate, Together AI and OpenRouter.

Source ↗
2026-10-15

Nemotron 3.5 Super VL + Bolt (NVIDIA)

NVIDIA released Nemotron 3.5 Super VL, an open 120B-parameter (12B active) multimodal model for always-on agents, plus Nemotron 3.5 Bolt 74B-A9B 24GB for local agent workloads on RTX and DGX Spark; Bolt E4B 4GB is due Oct 23.

Source ↗
2026-10-15

Scout 1.0 (Datadog)

Datadog introduced Scout 1.0, its first in-house model (post-trained from GLM-5.3 Flash with reinforcement learning on incident simulations), powering the Bits Investigation AI SRE agent at a fraction of frontier-model cost; in preview with signup open.

Source ↗
2026-10-15

Ternary Bonsai 2 2B / 4B / 9B (PrismML)

PrismML expanded its ternary-weight Bonsai 2 family with 2B, 4B and 9B models based on Qwen3.5, aimed at local agents on phones, laptops and other everyday devices; weights released under Apache 2.0, running via llama.cpp and MLX.

Source ↗
2026-10-15

Span-02 (Respan)

Respan released Span-02, an open-weight multimodal decision model with deadline-aware 'anytime' reasoning: give it a time budget and it returns its best structured answer so far (choices, yes/no, scores or numbers) with calibrated probabilities.

Source ↗
2026-10-15

Attie social vibe coding (Bluesky)

Bluesky expanded its AI app Attie so anyone can prompt-build websites, games, quizzes and polls that live on the AT Protocol and can be shared across Bluesky and the wider Atmosphere (announced at TechCrunch Disrupt 2026).

Source ↗
2026-10-15

OpenUI 1.0

The MIT-licensed generative UI framework from the thesysdev team reached 1.0: agents stream OpenUI Lang and render interactive components such as charts, forms and dashboards instead of walls of text; 1.0 adds custom functions and actions in the React runtime.

Source ↗
2026-10-15

Agent Vault (Self Labs)

Self Labs launched Agent Vault, a portable login and payment layer for AI agents: agents work with mock tokens while real card numbers and passwords are swapped in only at fill time, with spending limits, merchant allowlists and a kill switch; connects via MCP.

Source ↗
2026-10-15

Eney

MacPaw launched Eney, a proactive AI assistant for Mac and iPhone that runs multi-step tasks across local apps and connectors (Gmail, Notion, Slack, etc.), with a 14-day free credit trial.

Source ↗
2026-10-15

looot

looot launched on Product Hunt: one API key/balance for 2,500+ data APIs from 90+ providers for AI agents (MCP), with pre-call pricing and pay-per-run billing.

Source ↗
2026-10-14

HeyGen Voice

HeyGen released HeyGen Voice, a scene-aware voice-cloning model that adapts cloned speech to a video's room, mic distance, ambient sound and delivery style; it ranked #1 on Artificial Analysis' controlled-voice leaderboard (Oct 9 snapshot, Elo 1201).

Source ↗
2026-10-14

Pancake

Pancake launched as GTM infrastructure for AI agents (GTM Brain, enrichment, outreach Plays) via MCP for Claude/Codex; #1 on Product Hunt Oct 14.

Source ↗
2026-10-14

Inferock

Inferock launched accountable LLM inference (Watch / Serve / Dedicated) with per-call measurement and credits for eligible failed calls; #2 on Product Hunt Oct 14.

Source ↗
2026-10-14

Retriever AI

Retriever AI (rtrvr) launched a free AI browser agent (Chrome extension + cloud + API/MCP) for web tasks, outreach, and job applications; #3 on Product Hunt Oct 14.

Source ↗
2026-10-14

Tiden

Tiden launched intent verification for coding agents—maps product specs, ties promises to tests, and blocks closing sessions while promises remain open; works beside Claude Code, Cursor, Copilot, Codex.

Source ↗
2026-10-14

Tensormesh Platform

Tensormesh (creators of LMCache) launched Tensormesh Platform, enterprise KV-cache management for LLM inference across Kubernetes with tiered/prefix caching and multi-tenancy.

Source ↗
2026-10-14

Grok Voice 3.0

xAI launched Grok Voice 3.0, a speech-to-speech model ranked #1 on Artificial Analysis Speech-to-Speech Index, available in the Voice Agent API at $0.08/min.

Source ↗
2026-10-14

Deep Review

Google Research publicly launched Deep Review, an agentic document verification system (conference-piloted) with two free reviews/week and Gemini API key usage beyond that.

Source ↗
2026-10-14

Ultramarine 1 Flash

Paper Instruments released Ultramarine 1 Flash, a knowledge-work model (GLM-5.3 Flash + non-verifiable RL) with a live OpenAI-compatible API; docs at docs.paperinstruments.com.

Source ↗
2026-10-13

Drama 3 (Fish Audio)

Fish Audio launched Drama 3, an open-weight emotive TTS model with plain-language performance direction, #1 on Artificial Analysis Pronunciation Robustness, available on fish.audio, mobile apps, and API.

Source ↗
2026-10-13

OrcaRouter Agent Tools

OrcaRouter launched Agent Tools: a pay-per-call catalog of 1,400+ tools (search, scraping, SEO, enrichment) on one API key and workspace wallet, documented under Agent Infra.

Source ↗

Frequently asked questions

How many AI tools and models were released in October 2026?

SIloadout verified 107 AI tool, model and product launches in October 2026 (through October 30), across 3 weeks. The most recent was Drex Max v1 open weights (Nace AI) on 2026-10-30.

Which week of October 2026 had the most AI releases?

The week of Oct 19–25, 2026, with 45 verified launches in October 2026. By week: Oct 26–Nov 1, 2026: 29; Oct 19–25, 2026: 45; Oct 12–18, 2026: 33.

What was the latest AI release in October 2026?

Drex Max v1 open weights (Nace AI) (2026-10-30). Nace AI open-sourced Drex Max v1, a dense 27B multimodal decision model.

How does SIloadout verify releases?

Each launch is checked against the vendor's own post, its official X account or credible press before it is added, and every source link is confirmed to load. Rumored, unreleased or ambiguously dated items are left out.

← Newer: November 2026 · All releases · Atom feed

Browse by month: November 2026 (1) · October 2026 (107)