Cloud price beside a local guess
Cost board
API prices per million tokens, in pounds. The exchange rate is 0.74501 dollars to the pound on 2026-11-10, from the European Central Bank via Frankfurter. Prices were taken from OpenRouter on 2026-11-11.
Size class is our filter bucket, not a parameter count, unless the model id states one. Swipe the table sideways on a phone. 29 rows.
| Model | Input £ / M | Output £ / M | UK price | Weights | Provider | Size |
|---|---|---|---|---|---|---|
| Anthropic: Claude Haiku 4.5 | £0.745 | £3.73 | — | closed | Anthropic | small |
| Anthropic: Claude Opus 4.5 | £3.73 | £18.63 | — | closed | Anthropic | frontier |
| Anthropic: Claude Sonnet 4.5 | £2.24 | £11.18 | — | closed | Anthropic | frontier |
| DeepSeek: DeepSeek V3 | £0.238 | £0.663 | — | open | DeepSeek | frontier |
| DeepSeek: R1 0528 | £0.373 | £1.60 | — | open | DeepSeek | frontier |
| Google: Gemini 2.5 Flash | £0.224 | £1.86 | — | closed | medium | |
| Google: Gemini 2.5 Flash Lite | £0.0745 | £0.298 | — | closed | small | |
| Google: Gemini 2.5 Pro | £0.931 | £7.45 | — | closed | frontier | |
| Google: Gemma 3 12B | £0.0373 | £0.112 | — | open | small · 12B | |
| Google: Gemma 3 27B | £0.0596 | £0.335 | — | open | medium · 27B | |
| Meta: Llama 3.1 70B Instruct | £0.536 | £0.536 | — | open | Meta | large · 70B |
| Meta: Llama 3.1 8B Instruct | £0.0149 | £0.0298 | — | open | Meta | small · 8B |
| Meta: Llama 3.2 3B Instruct | £0.0373 | £0.246 | — | open | Meta | small · 3B |
| Meta: Llama 3.3 70B Instruct | £0.529 | £0.529 | — | open | Meta | large · 70B |
| Meta: Llama 4 Scout | £0.0745 | £0.224 | — | open | Meta | large |
| Microsoft: Phi 4 | £0.0522 | £0.104 | — | open | Microsoft | unspecified |
| Mistral: Ministral 3 8B 2512 | £0.112 | £0.112 | — | open | Mistral | small · 8B |
| Mistral: Mistral Nemo | £0.0216 | £0.0223 | — | open | Mistral | small |
| Mistral: Mistral Small 3.2 24B | £0.0745 | £0.224 | — | open | Mistral | medium · 24B |
| OpenAI: GPT-4.1 Mini | £0.298 | £1.19 | — | closed | OpenAI | small |
| OpenAI: GPT-4.1 Nano | £0.0745 | £0.298 | — | closed | OpenAI | small |
| OpenAI: GPT-4o-mini | £0.112 | £0.447 | — | closed | OpenAI | small |
| OpenAI: GPT-5 Mini | £0.186 | £1.49 | — | closed | OpenAI | small |
| OpenAI: GPT-5 Nano | £0.0373 | £0.298 | — | closed | OpenAI | small |
| OpenAI: gpt-oss-120b | £0.112 | £0.447 | — | open | OpenAI | large · 120B |
| OpenAI: gpt-oss-20b | £0.0134 | £0.0671 | — | open | OpenAI | medium · 20B |
| Qwen: Qwen3 32B | £0.0596 | £0.209 | — | open | Qwen | medium · 32B |
| Qwen: Qwen3 Coder 30B A3B Instruct | £0.112 | £0.447 | — | open | Qwen | medium · 30B-A3B |
| Qwen2.5 72B Instruct | £0.268 | £0.298 | — | open | Qwen | large · 72B |
Local electricity, when a GPU is chosen, uses the “chatbot for a small team” example and only a sourced speed. Where there is no sourced speed the cell says so. Wattage is the official graphics power, labelled an estimate, unless a run published watts. Price source.