Find the right model for the work.
Verified prices beside published capability evidence. Choose a task, then inspect the trade-off.
USE CASE
PRIORITY
Refine comparison
GENERAL CAPABILITY VIEW
48 scored modelsThe price–capability frontier.
| Model | Standard $ / 1M | ECI | Arena | Context |
|---|---|---|---|---|
 gpt-5-nanoFrontier | $0.138Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 139.3Epoch Capabilities Index90% interval 134.3–141.5 · 5 tested configurations · snapshot 2026-11-01.View source | — | 272k |
| gpt-6-lunaFrontier | $0.20Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 156.8Epoch Capabilities Index90% interval 154–159.5 · 3 tested configurations · snapshot 2026-11-01.View source | — | 922k |
| gpt-5.6-lunaFrontier | $0.45Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 156.7Epoch Capabilities Index90% interval 154.2–159.3 · 6 tested configurations · snapshot 2026-11-01.View source | — | 922k |
 Gemini 3.8 FlashFrontier | $1.50Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 157.6Epoch Capabilities Index90% interval 155.5–160.9 · 3 tested configurations · snapshot 2026-11-01.View source | — | 1,049k |
| gpt-6.1-solFrontier | $4.00Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 167.8Epoch Capabilities Index90% interval 164.3–172.5 · 4 tested configurations · snapshot 2026-11-01.View source | — | 922k |
 Claude Opus 5.5Frontier | $8.00Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 168.4Epoch Capabilities Index90% interval 165.4–173 · 5 tested configurations · snapshot 2026-11-01.View source | — | 1,000k |
| gpt-6-astraFrontier | $20.00Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 169.1Epoch Capabilities Index90% interval 165.9–174.3 · 5 tested configurations · snapshot 2026-11-01.View source | — | 922k |
| gpt-4o-mini | $0.262Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 126.5Epoch Capabilities Index90% interval 120.5–128.6 · 1 tested configuration · snapshot 2026-11-01.View source | — | 128k |
| gpt-5.4-nano | $0.463Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 146.1Epoch Capabilities Index90% interval 143.6–148.5 · 3 tested configurations · snapshot 2026-11-01.View source | — | 272k |
| Gemini 3.1 Flash-Lite | $0.563Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 144.3Epoch Capabilities Index90% interval 142.3–145.9 · 4 tested configurations · snapshot 2026-11-01.View source | — | 1,049k |
| gpt-5-mini | $0.688Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 145.4Epoch Capabilities Index90% interval 143.5–147.1 · 5 tested configurations · snapshot 2026-11-01.View source | — | 272k |
| gpt-4.1-mini | $0.70Direct API token tariff75% input / 25% output · verified 2026-11-01. Usage and reasoning effort can change the bill.View source | 134.9Epoch Capabilities Index90% interval 131.1–136.4 · 1 tested configuration · snapshot 2026-11-01.View source | — | 1,048k |
Showing 12 of 48 scored matches · 60 priced models passed the filters
How this view is calculated
Standard price = 0.75 × input + 0.25 × output, per 1M total tokens. ECI is Epoch AI’s general capability scale, often combining several effort settings. Arena measures relative human preference. Neither is a success probability. The frontier keeps models for which no other scored model has both a lower tariff and a higher ECI lower bound. For coding, Forcebench compares configurations on the same 272 tasks using source-estimated run costs. Missing task evidence falls back to general ECI; it is never called task performance.
Epoch ECIArena datasetForcebench