Fourteen models, one page. Every figure here is pulled from the same profile pages linked below — price, context window and licence, nothing invented for this table. Click any column heading to sort; click a row to open the full review.
Full comparison
Prices are per million tokens, published rate as of the “checked” date on each profile.
| Provider ↕ | Model ↕ | Input $/M ↕ | Output $/M ↕ | Context ↕ | Licence ↕ | Best for ↕ |
|---|---|---|---|---|---|---|
| Meta Superintelligence Labs | Muse Spark | Not published | Not published | Not disclosed | Proprietary | Everyday questions inside apps people already have open |
| Mistral AI | Mistral Medium 3.5 | $1.50 | $7.50 | 256K | Open weights | Coding agents running on a small GPU cluster |
| Moonshot AI | Kimi K3 | Not published | Not published | 1.05M | Open weights | Research teams that need frontier scale weights they can inspect |
| Alibaba Qwen | Qwen3.8-Max | $2.00 | $6.00 | 1M | Open weights | Multi day autonomous coding projects |
| DeepSeek | DeepSeek-V4-Flash | $0.44 | $1.32 | 1M | Open weights | Teams that need to run a capable model on their own infrastructure |
| xAI | Grok 4.5 | $2.00 | $6.00 | 500K | Proprietary | Terminal and agentic coding work where token efficiency matters |
| Google DeepMind | Gemini 3.1 Pro | $2.00 | $12.00 | 1.05M | Proprietary | Abstract reasoning problems and research inside the Google ecosystem |
| Google DeepMind | Gemini 3.6 Flash | $1.50 | $7.50 | 1M | Proprietary | Work that mixes documents, images, audio, and video in one request |
| Anthropic | Claude Fable 5 | $10.00 | $50.00 | 1M | Proprietary | Long horizon tasks where the cost of a wrong answer exceeds the cost of the tokens |
| Anthropic | Claude Sonnet 5 | $2.00 | $10.00 | 1M | Proprietary | Everyday reasoning and autonomous tool use without a subscription |
| Anthropic | Claude Opus 5 | $5.00 | $25.00 | 1M | Proprietary | Agents that have to keep going for hours without losing the thread |
| OpenAI | GPT-5.6 Luna | $0.20 | $1.20 | 1.05M | Proprietary | High volume work where cost per token decides the architecture |
| OpenAI | GPT-5.6 Terra | $2.00 | $12.00 | 1.05M | Proprietary | Everyday production work that still needs the full context window |
| OpenAI | GPT-5.6 Sol | $5.00 | $30.00 | 1.05M | Proprietary | Hard reasoning, agentic coding, and security research |
Input price vs. output price
Both axes are price per million tokens, log scale. We dropped context window as the second axis: eleven of the twelve priced models now sit within a 500K-to-1.05M-token band, so it barely spread the points apart. Output-to-input ratio does more work — it clusters into three bands by vendor tier (6×, 5×, 3×), visible here as three roughly parallel groups. Hover a point for detail; labels that would overlap are nudged clear with a leader line. Unpriced models are listed separately below.
What would this actually cost?
We’ve argued before that the number to compare is cost per completed task, not price per token. This is that argument made concrete: pick a volume of input text, see what running it through each priced model actually costs.