Ten frontier AI labs, side by side
What each one sells now, what it costs, where you can buy it and what it runs on. Each lab’s page links its figures to their sources.
01 · Compared
Flagship list prices
The facts, compared
As of 30 Oct 2026, five of the ten labs release their flagship with open weights. By input price, the cheapest flagship is DeepSeek-V4-Pro (GA), at $1.32 in and $3.96 out per million tokens, and the dearest are GPT-6 Astra and Claude Fable 5.1, each at $10 in and $50 out. The newest frontier release is Mistral Large 4 (open weights), from Mistral AI on 29 Oct 2026.
List prices are per million tokens for each lab’s flagship, standard tier, from its own pricing page. Sort by release date, price or releases.
| Lab | Price then | Weights | Where to buy | |||
|---|---|---|---|---|---|---|
| Mistral AI | Mistral Large 4 (open weights)29 Oct 2026 | $1.50 / $7.50Mistral Medium 3.5 | $0.40 / $2.00Mistral Medium 3 at launch, May 2025 | Open flagship | 5 placesMistral Studio, Hugging Face, Microsoft Foundry and Copilot Studio | 8 |
| OpenAI | GPT-6.1 Terra26 Oct 2026 | $10 / $50GPT-6 Astra | $1.25 / $10GPT-5, Sep 2025 | Closed flagship, some open | 5 placesOpenAI API, ChatGPT, Microsoft Foundry | 13 |
| Moonshot AI | Kimi K3.1 Pro24 Oct 2026 | $3.00 / $15.00Kimi K3 | $0.60 / $2.50Kimi K2 0905, Sep 2025 | Open flagship | 7 placesKimi API and Kimi apps, Hugging Face, Alibaba Cloud Model Studio | 5 |
| DeepSeek | DeepSeek-V4.1-Flash-101717 Oct 2026 | $1.32 / $3.96DeepSeek-V4-Pro (GA) | $0.56 / $1.68deepseek-chat, Sep 2025 | Open flagship | 6 placesDeepSeek API, DeepSeek app and web, Hugging Face | 6 |
| Meta | Muse Spark Open15 Oct 2026 | $4.00 / $20.00Muse Wonder / Muse Wonder Full | No earlier price | Closed flagship, some open | 4 placesmeta.ai and the Meta AI app, Muse personal agent, Meta Model API | 8 |
| Gemini 4 Argon (GA)13 Oct 2026 | $2.00 / $10.00Gemini 4 Argon (GA) | $1.25 / $10.00Gemini 2.5 Pro, Sep 2025 | Closed flagship, some open | 6 placesGemini API and Google AI Studio, Gemini Enterprise Agent Platform, Gemini Enterprise | 9 | |
| Anthropic | Claude Sonnet 5.528 Sep 2026 | $10 / $50Claude Fable 5.1 | $15 / $75Claude Opus 4.1, Sep 2025 | Closed | 7 placesClaude API, Amazon Bedrock, Google Cloud | 14 |
| xAI | Grok 4.721 Sep 2026 | $2.00 / $6.00Grok 4.7 | $3.00 / $15.00Grok 4, Sep 2025 | Closed flagship, some open | 5 placesxAI API, Cursor and Grok Build, Oracle Cloud Infrastructure | 7 |
| Alibaba (Qwen) | Qwen3.8-Max-09022 Sep 2026 | $2.00 / $6.00Qwen3.8-Max | $1.20 / $6.00qwen3-max, Sep 2025 | Open flagship | 5 placesQwen Cloud and Model Studio, Bailian, Qwen app | 11 |
| Zhipu (Z.ai) | GLM-5.3-Flash26 Aug 2026 | $1.40 / $4.40GLM-5.3 | $0.60 / $2.20GLM-4.6, Sep 2025 | Open flagship | 6 placesz.ai and bigmodel.cn, GLM Coding Plan, Hugging Face | 7 |
02 · Every frontier release
175 releases
Ten labs shipped 75 frontier releases in the year to October 2026, up from 53
- OpenAI 23 GPT-6.1 Terra26 Oct 2026
- Anthropic 21 Claude Sonnet 5.528 Sep 2026
- Google 19 Gemini 4 Argon (GA)13 Oct 2026
- DeepSeek 19 DeepSeek-V4.1-Flash-101717 Oct 2026
- Meta 11 Muse Spark Open15 Oct 2026
- xAI 14 Grok 4.721 Sep 2026
- Mistral 17 Mistral Large 4 (open weights)29 Oct 2026
- Alibaba 27 Qwen3.8-Max-09022 Sep 2026
- Moonshot 9 Kimi K3.1 Pro24 Oct 2026
- Zhipu 15 GLM-5.3-Flash26 Aug 2026
One dot per frontier release at its date. The full list, with sources and downloads, is on the release chart. Which plans and apps the newest OpenAI, Google, SpaceXAI and Anthropic models reached, and when, is in The Access Gap.
03 · Our read, not a fact
opinion
Where most of each lab’s value sits
The question we ask of every lab: is most of its value in the model, which gets cheaper fastest, or in something other work is built on? Each lab’s page says what would change our read.
In what others build on
Most of the value sits in something other work is built on: a platform, a chip stack, a distribution channel.
Split
The value is split between the model and what others build on.
In the model
Most of the value sits in the model itself, the layer that gets cheaper fastest.