Economics
Price per token is not price per finished work
$/1M input and output stay listed. The numbers that should decide a purchase are $/completed task, $/correct task, tokens/task, and effective work rate.
Universe index vs $ per correct task
Upper-left is rare: high capability at low cost-per-success. Most frontier models sit right and high.
| Endpoint | $ / 1M in | $ / 1M out | $ / task | $ / correct | Tokens / task | Tools / task | EWR | Evidence |
|---|---|---|---|---|---|---|---|---|
| Llama 4 70B Groq LPU | $0.20 | $0.20 | $0.0041 | $0.0071 | 4,720 | 2.0 | 64.3 | LIVE_MEASURED |
| Llama 4 70B Together | $0.18 | $0.18 | $0.0038 | $0.0064 | 4,720 | 2.0 | 32.3 | LIVE_MEASURED |
| Gemini 2.5 Flash Google API | $0.15 | $0.60 | $0.0090 | $0.013 | 4,980 | 2.2 | 20.2 | LIVE_MEASURED |
| Grok 3 Mini xAI API | $0.30 | $0.50 | $0.0084 | $0.013 | 4,700 | 2.1 | 16.6 | LIVE_MEASURED |
| Llama 4 405B Self-host B200 | $0.28 | $0.28 | $0.0078 | $0.012 | 5,350 | 2.6 | 13.5 | REPRODUCED |
| GPT-5 Mini OpenAI API | $0.40 | $1.60 | $0.014 | $0.020 | 4,950 | 2.4 | 8.50 | LIVE_MEASURED |
| Llama 4 405B Groq LPU | $0.70 | $0.90 | $0.020 | $0.030 | 5,350 | 2.5 | 8.05 | LIVE_MEASURED |
| Llama 4 405B Self-host H100 | $0.40 | $0.40 | $0.011 | $0.016 | 5,350 | 2.6 | 4.76 | REPRODUCED |
| DeepSeek V3.1 DeepSeek API | $0.27 | $1.10 | $0.018 | $0.024 | 5,700 | 2.7 | 4.52 | LIVE_MEASURED |
| Llama 4 405B Fireworks | $0.90 | $0.90 | $0.024 | $0.036 | 5,350 | 2.6 | 3.43 | LIVE_MEASURED |
| Llama 4 405B Together | $0.80 | $0.80 | $0.022 | $0.032 | 5,350 | 2.6 | 2.97 | LIVE_MEASURED |
| Qwen3-Max Together | $0.40 | $1.60 | $0.026 | $0.036 | 5,700 | 2.8 | 2.56 | LIVE_MEASURED |
| Gemini 2.5 Pro Google API | $1.25 | $10.00 | $0.041 | $0.053 | 5,800 | 2.9 | 1.96 | LIVE_MEASURED |
| Mistral Large 3 Mistral API | $2.00 | $6.00 | $0.048 | $0.068 | 5,200 | 2.9 | 1.32 | LIVE_MEASURED |
| Grok 4 xAI API | $3.00 | $15.00 | $0.074 | $0.095 | 5,300 | 3.2 | 0.925 | LIVE_MEASURED |
| Claude 4.1 Sonnet Anthropic API | $3.00 | $15.00 | $0.082 | $0.108 | 6,300 | 3.4 | 0.724 | LIVE_MEASURED |
| GPT-5 OpenAI API | $5.00 | $20.00 | $0.210 | $0.262 | 5,700 | 3.0 | 0.268 | LIVE_MEASURED |
| GPT-5 Azure OpenAI | $5.50 | $22.00 | $0.232 | $0.292 | 5,700 | 3.0 | 0.217 | LIVE_MEASURED |
| Grok 4 Heavy xAI API | $5.00 | $25.00 | $0.168 | $0.205 | 6,900 | 2.8 | 0.198 | LIVE_MEASURED |
| Claude 4.1 Opus Anthropic API | $15.00 | $75.00 | $0.420 | $0.519 | 6,700 | 3.8 | 0.104 | LIVE_MEASURED |
| o4-pro OpenAI API | $15.00 | $60.00 | $0.620 | $0.738 | 10,000 | 2.2 | 0.033 | LIVE_MEASURED |