AI models: prices, context and release dates
14 models with the figures we hold and where each came from. No benchmark scores and no ranking: those need evidence we have not collected yet.
Every tracked model
Alphabetical. Order does not imply quality. Open a model for its sources and what we do not know.
All models
| Model | Provider | Weights | Context | Input, $ per 1M | Output, $ per 1M |
|---|---|---|---|---|---|
| Claude Fable 5.1 | Anthropic | Closed | 1M | $10 | $50 |
| Claude Opus 5.5 | Anthropic | Closed | 1M | $4 | $20 |
| Claude Sonnet 5.5 | Anthropic | Closed | 1M | $2 | $10 |
| DeepSeek V4.1 Flash | DeepSeek | Open | 1M | $0.30 | $1.20 |
| Gemini 3.8 Flash | Closed | Not supplied | $0.75 | $3.75 | |
| GLM-5.3 | Zhipu (Z.ai) | Open | Not supplied | $1.40 | $4.40 |
| GPT-6 Astra | OpenAI | Closed | 1.05M | $10 | $50 |
| GPT-6 Luna | OpenAI | Closed | 1.05M | $0.10 | $0.50 |
| GPT-6.1 Sol | OpenAI | Closed | 1.05M | $2 | $10 |
| Grok 4.7 | xAI | Closed | 500K | $2 | $6 |
| Kimi K3 | Moonshot AI | Open | 1M | Not supplied | Not supplied |
| Mistral Medium 3.5 | Mistral | Closed | Not supplied | Not supplied | Not supplied |
| Muse Spark 1.3 | Meta | Closed | Not supplied | Not supplied | Not supplied |
| Qwen3.8 Max | Alibaba/Qwen | Closed | Not supplied | Not supplied | Not supplied |
4 of 14 models have no supplied price yet. Their pages stay out of search results until they do.
Change log for model figures
Every price, context or status change, dated. Nothing is listed here that we did not see change.
Change log
- Baseline. 14 models recorded with list prices, context windows and release dates as supplied by the editorial team. None re-verified against vendor pages yet. No price or context change has been observed since, so none is listed.
Frequently asked questions
Where do these model figures come from?
The editorial team supplied them on 2026-10-14. Each model page links to the provider's documentation and pricing page so you can check. We have not yet re-verified the figures ourselves, and we say so wherever they appear.
Why are there no benchmark scores?
A score must meet the requirements on our benchmarks page before we publish it, and none has. Vendor-reported numbers on mixed scales are not comparable, so we publish none rather than imply a ranking.
Why is a model page marked noindex?
A model with no supplied price would be a page of unknowns. We keep it out of search results until a price is supplied, but you can still open it from this page.
How do I choose between these models?
Use the Model Finder to shortlist models that fit your task, inputs, context size and budget, then test the shortlist on your own prompts. The cost calculator shows what each would cost at your volume.