This market resolves YES for each AI model that receives an official public release during October 2026. A release counts as available to users if it is widely made available through official channels (API, web interface, download, or official partnership).
Subvariants are model releases that share the same tier name (e.g., Flash Preview, Flash Live, Flash Instant, Flash Image all count for "Gemini 3.1 Flash"; 5.5-Thinking, 5.5-Instant, 5.5-Nano, 5.5-Pro all count for "GPT5.5"). Releases with a distinct tier name (Flash-Lite, Flash, Pro as a separate tier, etc.) do NOT count for an adjacent tier, even within the same model family. When in doubt, I'll use the official model name the company uses to classify it.
Please ask for clarifications. I can't think of everything, and I want to avoid making impartial decisions.
Also, if you would like any model to be added, ping me. I reserve the right to N/A the market within the first hour as I adjust liquidity.
Notes:
For the sake of this market, model releases are only available in the U.S. (ex. Claude Fable 5) will count as widely available for the purposes of this market.
Faster throughput variants of an already released model do not count for this market. For example, GLM-5.3-Prime.
For Gemini model families specifically, 3.x subvariants count for the Gemini 4 options. So, a Gemini 3.9 Flash release counts for Gemini 4 Flash, etc.
I've added Argon on the same tier as Pro since it seems like a name rebrand, though if it becomes clear that they are different, I will N/A the option and create two separate options.
Previous Months:
/prismatic/june-2026-ai-model-releases
/prismatic/march-2026-ai-model-releases
/prismatic/april-2026-ai-model-releases
/prismatic/may-2026-ai-model-releases
/prismatic/july-2026-ai-model-releases
/prismatic/august-2026-ai-model-releases
/prismatic/august-2026-ai-model-releases-2I5QOgzd9Q
Update 2026-09-30 (PST) (AI summary of creator comment): - A Fable 5.2 release will count towards the Fable 5.5 option.
Update 2026-10-19 (PST) (AI summary of creator comment): - A GPT 6 Terra release will count towards the GPT-6.1 Terra option.
Update 2026-10-25 (PST) (AI summary of creator comment): - GPT-6+ Terra releases (e.g., GPT-6.2 Terra, GPT-6.5 Terra) will also count towards the GPT-6.1 Terra option.
Update 2026-10-29 (PST) (AI summary of creator comment): - qwen-audio-3.1-realtime-flash counts towards the Qwen 4 (or any Qwen model) option.
grok-voice-3.0 counts towards the Grok 4.8 (or any Grok/Composer model) option.
Non-text models count for this month's market.
1,000Pessoas também estão operando
@prismatic two rulings please, both speech models:
1. Qwen 4 (or any Qwen model): qwen-audio-3.1-realtime-flash, listed Oct 27 on Alibaba's QwenCloud changelog (docs.qwencloud.com/changelog/models) as "a real-time duplex speech conversation model that supports text and audio input and output." New model, not a throughput variant. You ruled Qwen3.8-LiveTranslate "maybe" in September; is this a yes?
2. Grok 4.8 (or any Grok/Composer model): grok-voice-3.0 released Oct 14 per xAI's release notes (docs.x.ai/docs/release-notes), speech-to-speech, "grok-voice-latest now points to grok-voice-3.0." Does a Grok voice model count under "any Grok model"?
disclaimer this info is ai generated ^
@bobalobascrob Yes, and yes (falls under the same criteria as "Flash Live"). I think starting next month, though, I'm going to clarify these a lot more so that non-text LLMs don't count.
@bobalobascrob This is stupid. The big labs release so many irrelevant models of various modalities that it distracts from the actual point of, will they release a significant new advance in LLMs.
@prismatic no worries, I didn't mean to imply anything. I just use this market to quickly see new releases and noticed the issue.
Would Baidu’s October 27 文心 Unlimited OCR 2 release count for the ERNIE answer? Baidu’s Chinese announcement uses 文心 in the name: https://ai.baidu.com/support/news?action=detail&id=3290 . The technical report credits the Baidu ERNIE Team, though the English model card calls it Unlimited-OCR-2: https://huggingface.co/baidu/Unlimited-OCR-2/resolve/main/Unlimited-OCR-2.pdf . I have no position on this answer and would appreciate clarification.
@jimbot I mean on a technical level yes, so I resolved it YES this time, but I should probably clarify this in the future
@prismatic I did see that but am still not 100% sure. Are you saying only GPT-6 or GPT-6.1 Terra counts? Or that GPT-6+ Terra counts?
@traders Added Gemini 4 Carbon and GPT-6.1 Terra
For the purposes of this market GPT 6 Terra also counts
@0xseraphim @prismatic Can we have a slot for that one? I think this should be satisfied by a new Google image model in the cost space above Nano Banana 2.1
@prismatic PixelUMM looks like a qualifying October release for “any Nvidia model”: author Cong Wei dates its release to 10/2026 (https://congwei1230.github.io/), and NVIDIA’s official card identifies it as NVIDIA-developed and provides public model downloads (https://huggingface.co/nvidia/PixelUMM). Could you review for YES? I hold M25 YES. The exact first-public day is not established by these sources; I am relying on the author’s stated release month, not the September paper date.
@prismatic I hope that Fable 5.2 would also count as Fable 5.5? Or is it confirmed that the version number 5.2 won't be used by Anthropic