Code generation: a model at 1/271st the price of the best scorer, at 97.8% of its quality.

Model pricing, measured

Per-token list prices next to the number a bill is actually made of: live-measured cost per 1,000 requests, per kind of work — hidden reasoning tokens included.

modelvendor$ / 1M in$ / 1M outmeasured $ / 1K requests
claude-haiku-4.5anthropic$1.00$5.00$1.12
claude-sonnet-4.5anthropic$3.00$15.00$3.31 – $7.66
gemini-2.5-flashgoogle$0.30$2.50$0.0391 – $0.9105
gpt-4.1-miniopenai$0.40$1.60$0.1081 – $0.2520

only models on a live measured frontier appear · measured ranges span the workloads each model earns · method

Model pricing, measured per request · Frontier Notes