Cost Comparison by Model
Per-token cost across providers for your task type and volume.
Showing per-model monthly cost across all 23 Prism catalog models for the Simple task. Cheapest model: Llama 3.1 8B (Groq) at $6.40/mo. Most expensive: Claude Opus 4.1 (legacy) at $3.5K/mo (539x cheaper-to-priciest spread).
| Rank | Model | Provider | Cost / req | Monthly | Prism pick |
|---|---|---|---|---|---|
| 1 | Llama 3.1 8B (Groq) | Groq | $0.000064 | $6.40 | eco |
| 2 | Llama 3.1 8B (Cerebras) | Cerebras | $0.000110 | $11 | |
| 3 | Gemini 2.5 Flash | $0.000150 | $15 | ||
| 4 | groq-gpt-oss-20b | Groq | $0.000150 | $15 | |
| 5 | groq-gpt-oss-safeguard-20b | Groq | $0.000150 | $15 | |
| 6 | Ministral 8B | Mistral | $0.000165 | $17 | |
| 7 | mistral-small-3-2 | Mistral | $0.000170 | $17 | |
| 8 | Llama 4 Scout (Groq) | Groq | $0.000190 | $19 | |
| 9 | DeepSeek V4 Flash | DeepSeek | $0.000196 | $20 | |
| 10 | fireworks-deepseek-v4-flash | Fireworks | $0.000196 | $20 | |
| 11 | gemini-2-5-flash-lite | $0.000200 | $20 | ||
| 12 | GPT-4o mini | OpenAI | $0.000300 | $30 | |
| 13 | gpt-oss 120B (Groq) | Groq | $0.000300 | $30 | |
| 14 | groq-gpt-oss-120b | Groq | $0.000300 | $30 | |
| 15 | mistral-small-4 | Mistral | $0.000300 | $30 | |
| 16 | Qwen3 32B (Groq) | Groq | $0.000409 | $41 | |
| 17 | cerebras-gpt-oss-120b | Cerebras | $0.000505 | $51 | |
| 18 | Codestral | Mistral | $0.000510 | $51 | |
| 19 | fireworks-minimax-m2p7 | Fireworks | $0.000600 | $60 | |
| 20 | fireworks-minimax-m2p5 | Fireworks | $0.000600 | $60 | |
| 21 | gemini-3-1-flash-lite | $0.000650 | $65 | ||
| 22 | Llama 3.3 70B (Groq) | Groq | $0.000709 | $71 | |
| 23 | Qwen 235B (Cerebras) | Cerebras | $0.000840 | $84 | |
| 24 | Mistral Large | Mistral | $0.000850 | $85 | |
| 25 | mistral-large-3 | Mistral | $0.000850 | $85 | |
| 26 | magistral-small | Mistral | $0.000850 | $85 | |
| 27 | devstral-2 | Mistral | $0.000920 | $92 | |
| 28 | gemini-2-5-flash | $0.000990 | $99 | ||
| 29 | fireworks-kimi-k2p5 | Fireworks | $0.001380 | $138 | |
| 30 | groq-kimi-k2-instruct-0905 | Groq | $0.001700 | $170 | |
| 31 | GPT-5.4 Mini | OpenAI | $0.001950 | $195 | |
| 32 | Kimi K2 (Fireworks) | Fireworks | $0.001960 | $196 | |
| 33 | fireworks-kimi-k2p6 | Fireworks | $0.001960 | $196 | |
| 34 | Claude Haiku 4 | Anthropic | $0.002300 | $230 | |
| 35 | Claude Haiku 4.5 | Anthropic | $0.002300 | $230 | |
| 36 | DeepSeek V4 Pro | DeepSeek | $0.002436 | $244 | |
| 37 | fireworks-deepseek-v4-pro | Fireworks | $0.002436 | $244 | |
| 38 | GLM 5.1 (Fireworks) | Fireworks | $0.002440 | $244 | |
| 39 | cerebras-zai-glm-4-7 | Cerebras | $0.002625 | $263 | |
| 40 | Magistral Medium | Mistral | $0.003100 | $310 | |
| 41 | Mistral Medium | Mistral | $0.003450 | $345 | |
| 42 | mistral-medium-3-5 | Mistral | $0.003450 | $345 | |
| 43 | Gemini 3.5 Flash | $0.003900 | $390 | ||
| 44 | Gemini 2.5 Pro | $0.004000 | $400 | ||
| 45 | gemini-2-5-pro | $0.004000 | $400 | ||
| 46 | GPT-4o | OpenAI | $0.005000 | $500 | |
| 47 | gemini-3-1-pro-preview | $0.005200 | $520 | ||
| 48 | Gemini 3.5 Pro (alias → 3.1 Pro) | $0.005200 | $520 | ||
| 49 | GPT-5.4 | OpenAI | $0.006500 | $650 | |
| 50 | Claude Sonnet 4 | Anthropic | $0.006900 | $690 | |
| 51 | Claude Sonnet 4.7 | Anthropic | $0.006900 | $690 | |
| 52 | Claude Sonnet 4.6 | Anthropic | $0.006900 | $690 | |
| 53 | Claude Sonnet 4.5 | Anthropic | $0.006900 | $690 | |
| 54 | Claude Opus 4.7 | Anthropic | $0.0115 | $1.1K | |
| 55 | Claude Opus 4.6 | Anthropic | $0.0115 | $1.1K | |
| 56 | Claude Opus 4.5 | Anthropic | $0.0115 | $1.1K | |
| 57 | GPT-5.5 | OpenAI | $0.0130 | $1.3K | |
| 58 | Claude Opus 4 | Anthropic | $0.0345 | $3.5K | sport |
| 59 | Claude Opus 4.1 (legacy) | Anthropic | $0.0345 | $3.5K |
Cost = (input_tokens × input_price + output_tokens × output_price) × volume. Prices are USD per 1M tokens, current Prism catalog (updated 2026-05-24). Highlighted models are what Prism's router picks for this task type in each mode (eco / balanced / sport). Cached requests cost $0 — model with the cache-hit-rate estimator for the full picture.
Frequently asked questions
- Is Cost Comparison by Model really free?
- Yes — fully free, no email gate, no signup, no sales call. All Prism tools live on this page and similar dedicated tool pages.
- What data does Cost Comparison by Model use?
- The calculations use current public provider pricing (Anthropic, OpenAI, Google) and Prism's published cache-economics math from production traffic. Nothing about your usage is captured — the calculator runs entirely in your browser.