Latest AI launches
Newest first, checked 25 Oct 2026. Filter in your browser with the buttons below — this is a static site, so filtering runs on your device, no account needed.
Showing 16 of 16 launches
Kimi K3.1 Pro — Moonshot's open-weight flagship, service first, weights to follow
The most recent flagship-class release tracked in October: general, coding and multimodal work from Moonshot AI.
Access: General · Price: Not published in the sources checked
Carbon-X — Hugging Face open model that chats with DNA
An open 14.8-billion-parameter model that takes raw DNA in a chat message and answers in English — for researchers, not for medical decisions.
Access: Open-weight · Price: Free to download, Apache 2.0 licence; run it yourself, no API price published
FLUX 3 Max — video model, price and access not yet published
A video-generation release in the same 19 Oct cluster as MiniMax's flash model.
Access: Not published · Price: Not published in the sources checked
MiniMax M3.1 Flash — fast tier, price not yet published
A speed-focused release: general use and coding, positioned as the fast tier in MiniMax's line.
Access: Not published · Price: Not published in the sources checked
Muse Spark Open — Meta's open-weight general model, licence to confirm
Meta's open release on 15 Oct, for general, coding and multimodal use.
Access: Open-weight · Price: Not published in the sources checked
FLUX 3 Image Klein — open image model for ordinary graphics cards
The image sibling of FLUX 3 Max, released the same week, in open form.
Access: Open-weight · Price: Not published in the sources checked
NVIDIA Nemotron 3.5 Super VL — open model for specialised work, claims to check
NVIDIA's open Nemotron 3.5 Super VL, reported as beating Claude Opus 5 and GPT-6.1 Sol on specialised task benchmarks in VentureBeat's coverage.
Access: Open-weight · Price: Not published in the sources checked
Gemini 4 Argon — agentic flagship for paid developers, price doubles 1 Dec
Google opened Gemini 4 Argon to paid Gemini API developers on 13 Oct, after a 30 Sep defender preview.
Access: General · Price: Intro $2 input / $10 output per 1M tokens to 30 Nov 2026, then $4 / $20
Muse Wonder and Wonder Full — weaker model by API, stronger by subscription only
Meta is directing users from Meta AI toward Muse Wonder. Only the weaker model is available via API; the strongest is subscription-only so far.
Access: API + Subscription · Price: No pay-as-you-go API price announced for Wonder Full; available via two Muse Code subscription tiers
GPT-6 in ChatGPT — Sol for paid, Luna for free, Astra for hard reasoning
OpenAI's GPT-6 rollout in ChatGPT from 7 Oct 2026: Sol for paid users, Luna for free users (later Help pages cast free access into doubt — check your account), Astra for complex reasoning and coding.
Access: General · Price: Sol $2 / $10 per 1M tokens; Luna $0.10 / $0.50 per 1M tokens; Astra price not published in sources checked
Claude Haiku 5.5 — fastest small model, priced to match GPT-6 Luna
Anthropic's fastest small model, priced to match OpenAI's cheapest tier, for summarising, classifying and database work.
Access: General · Price: $0.10 input / $0.50 output per 1M tokens for requests under 100,000 tokens
MAI-Code-1.1-Flash — fast coding model, caps and price not published
A fast coding model from Microsoft AI, released the same day as Haiku 5.5.
Access: Not published · Price: Not published in the sources checked
Mistral Large 4 — public preview now, weights promised by end of October
Mistral's large general model, with multimodal and reasoning coverage, released 6 Oct.
Access: Preview · Price: $1.36 / $4.18 per 1M tokens posted; $0.68 / $2.09 in public preview
Nano Banana 2.1 — image model, changes not yet published here
Google DeepMind's image/multimodal release on 6 Oct, distinct from Gemini 4 Argon (language) a week later.
Access: Not published · Price: Not published in the sources checked
EmbeddingGemma 2 — embedding model for search, not a chatbot
A specialist release: embeddings for search and retrieval, not a chatbot.
Access: Not published · Price: Not published in the sources checked
Beam — general and reasoning model, details not yet published
Reflection AI's general/reasoning/coding release on 5 Oct.
Access: Not published · Price: Not published in the sources checked
Next: launch timeline · coming next