What your workload costs a month
Pick a job or type your own token counts. Every text model we track, cheapest first, with whether it passes our test for that job.
At this volumeThe cheapest model that passes our test is GLM-5.3-Flash at $4.25 a month.
| # | Model | Monthly cost | Score | Passes our test |
|---|---|---|---|---|
| 1 | Claude Haiku 5.5 · Anthropic | $3.50 | 34 | No |
| 2 | GPT-6 Luna · OpenAI | $3.50 | 31 | No |
| 3 | Qwen3.8 Flash · Alibaba (Qwen) | $4.13 | – | Not scored |
| 4 | GLM-5.3-Flash · Z.ai | $4.25 | 67 | Yes |
| 5 | Nemotron 3.5 Super VL · NVIDIA | $4.25 | 34 | No |
| 6 | Gemma 4 31B IT · Google | $4.60 | 51 | No |
| 7 | Mistral Small 4 · Mistral AI | $4.65 | – | Not scored |
| 8 | MiniMax-M3.1-Flash · MiniMax | $4.65 | – | Not scored |
| 9 | DeepSeek V4.1 Flash 1017 · DeepSeek | $9.30 | – | Not scored |
| 10 | MiniMax-M3 · MiniMax | $9.30 | 49 | No |
| 11 | Gemini 3.5 Flash Lite · Google | $14.50 | 46 | No |
| 12 | Nova 2 Lite · Amazon | $15.95 | 0 | No |
| 13 | Nemotron 3 Ultra 550B A55B · NVIDIA | $18.60 | 38 | No |
| 14 | Qwen3.6 27B · Alibaba (Qwen) | $23.40 | – | Not scored |
| 15 | Gemini 3.8 Flash · Google | $26.25 | 78 | Yes |
| 16 | DeepSeek V4 Pro · DeepSeek | $35.64 | 59 | Yes |
| 17 | Muse Spark 1.3 · Meta | $35.75 | 71 | Yes |
| 18 | Mistral Large 4 · Mistral AI | $37.12 | 46 | No |
| 19 | GLM-5.3 · Z.ai | $38.60 | 67 | Yes |
| 20 | Mistral Medium 3.5 · Mistral AI | $52.50 | 40 | No |
| 21 | Grok 4.7 · xAI | $54.00 | 31 | No |
| 22 | Qwen3.8 Max · Alibaba (Qwen) | $54.00 | 72 | Yes |
| 23 | Claude Sonnet 5.5 · Anthropic | $70.00 | 69 | Yes |
| 24 | Claude Sonnet 5 · Anthropic | $70.00 | 56 | Yes |
| 25 | GPT-6.1 Sol · OpenAI | $70.00 | 58 | Yes |
| 26 | GPT-6 Sol · OpenAI | $70.00 | 37 | No |
| 27 | Gemini 4 Argon · Google | $70.00 | 97 | Yes |
| 28 | Kimi K3.1 Pro · Moonshot AI | $70.00 | – | Not scored |
| 29 | Command A Plus · Cohere | $77.50 | – | Not scored |
| 30 | GPT-5.6 Terra · OpenAI | $78.00 | 57 | Yes |
| 31 | Kimi K3 · Moonshot AI | $105 | 73 | Yes |
| 32 | Sonar Pro · Perplexity | $105 | – | Not scored |
| 33 | Claude Opus 5.5 · Anthropic | $140 | 90 | Yes |
| 34 | GPT-5.6 Sol · OpenAI | $140 | 66 | Yes |
| 35 | Muse Wonder · Meta | $140 | 79 | Yes |
| 36 | Claude Opus 5 · Anthropic | $175 | 84 | Yes |
| 37 | Claude Fable 5.1 · Anthropic | $350 | 83 | Yes |
| 38 | GPT-6 Astra · OpenAI | $350 | 56 | Yes |
List prices per million tokens, mostly from models.dev and checked against OpenRouter where we can. Cached input uses each model’s listed cache price; where none is listed we charge full input price. Scores are provisional. See the methodology.