AN
Anthropic API
Anthropic · first-party · est. 2021
Anthropic's first-party API. Canonical reference for all Claude models.
first-party frontier managed
Attested runs
44
on this provider
Models
8
served
Benchmarks
21
covered
Avg drift
0.0pp
vs canonical (n=0)
Default quant
FP-BF16
precision
Attested models on this provider
| Model | Benchmark | Anthropic API | Canonical | Drift |
|---|---|---|---|---|
| claude-3-7-sonnet-20250219 Self-reported | gpqa | 68.0 | ★ self | — |
| claude-3-7-sonnet-20250219 Self-reported | humaneval | 93.0 | ★ self | — |
| claude-3-7-sonnet-20250219 Self-reported | swe-bench-verified | 62.3 | ★ self | — |
| claude-haiku-4-5-20251001 Self-reported | gpqa | 51.8 | ★ self | — |
| claude-haiku-4-5-20251001 Self-reported | humaneval | 85.6 | ★ self | — |
| claude-haiku-4-5-20251001 Self-reported | math | 61.2 | ★ self | — |
| claude-haiku-4-5-20251001 Self-reported | mbpp | 82.4 | ★ self | — |
| claude-haiku-4-5-20251001 Self-reported | mmlu-pro | 66.7 | ★ self | — |
| claude-haiku-4-5 Attested | gsm8k | 100.0 ±16 | ★ self | — |
| claude-haiku-4.5 Attested | arc-challenge | 87.5 ±22 | ★ self | — |
| claude-haiku-4.5 Attested | commonsenseqa | 87.5 ±22 | ★ self | — |
| claude-haiku-4.5 Attested | gsm8k | 100.0 ±16 | ★ self | — |
| claude-haiku-4.5 Attested | hellaswag | 100.0 ±16 | ★ self | — |
| claude-haiku-4.5 Attested | math-500 | 37.5 ±28 | ★ self | — |
| claude-haiku-4.5 Attested | mmlu | 87.5 ±22 | ★ self | — |
| claude-haiku-4.5 Attested | mmlu-pro | 50.0 ±28 | ★ self | — |
| claude-haiku-4.5 Attested | openbookqa | 100.0 ±16 | ★ self | — |
| claude-haiku-4.5 Attested | truthfulqa | 100.0 ±16 | ★ self | — |
| claude-haiku-4.5 Attested | winogrande | 87.5 ±22 | ★ self | — |
| claude-opus-4-7 Self-reported | aime | 39.5 | ★ self | — |
| claude-opus-4-7 Self-reported | arc-agi | 21.0 | ★ self | — |
| claude-opus-4-7 Self-reported | bigcodebench | 47.2 | ★ self | — |
| claude-opus-4-7 Self-reported | gaia | 39.4 | ★ self | — |
| claude-opus-4-7 Self-reported | gpqa | 75.8 | ★ self | — |
| claude-opus-4-7 Attested | gsm8k | 100.0 ±16 | ★ self | — |
| claude-opus-4-7 Self-reported | hellaswag | 95.4 | ★ self | — |
| claude-opus-4-7 Attested | humaneval | 0.0 ±16 | ★ self | — |
| claude-opus-4-7 Self-reported | livecodebench | 54.0 | ★ self | — |
| claude-opus-4-7 Self-reported | math | 76.3 | ★ self | — |
| claude-opus-4-7 Attested | math-500 | 37.5 ±28 | ★ self | — |
| claude-opus-4-7 Self-reported | mbpp | 92.1 | ★ self | — |
| claude-opus-4-7 Self-reported | mmlu | 91.4 | ★ self | — |
| claude-opus-4-7 Self-reported | mmlu-pro | 85.2 | ★ self | — |
| claude-opus-4-7 Self-reported | swe-bench-verified | 64.3 | ★ self | — |
| claude-opus-4-7 Self-reported | tau-bench | 62.7 | ★ self | — |
| claude-opus-4-7 Self-reported | truthfulqa | 67.4 | ★ self | — |
| claude-opus-4.7 Attested | arc-challenge | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | commonsenseqa | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | gsm8k | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | hellaswag | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | math-500 | 37.5 ±28 | ★ self | — |
| claude-opus-4.7 Attested | mmlu | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | mmlu-pro | 62.5 ±28 | ★ self | — |
| claude-opus-4.7 Attested | openbookqa | 87.5 ±22 | ★ self | — |
| claude-opus-4.7 Attested | truthfulqa | 100.0 ±16 | ★ self | — |
| claude-opus-4.7 Attested | winogrande | 87.5 ±22 | ★ self | — |
| claude-sonnet-4-5-20250929 Self-reported | gpqa | 66.3 | ★ self | — |
| claude-sonnet-4-5-20250929 Attested | gsm8k | 94.1 ±12 | ★ self | — |
| claude-sonnet-4-5-20250929 Attested | humaneval | 91.5 ±13 | ★ self | — |
| claude-sonnet-4-5-20250929 Self-reported | math | 70.4 | ★ self | — |
| claude-sonnet-4-5-20250929 Attested | mbpp | 88.2 ±14 | ★ self | — |
| claude-sonnet-4-5-20250929 Self-reported | mmlu | 89.7 | ★ self | — |
| claude-sonnet-4-5-20250929 Self-reported | mmlu-pro | 78.4 | ★ self | — |
| claude-sonnet-4-5-20250929 Self-reported | swe-bench-verified | 49.0 | ★ self | — |
| claude-sonnet-4.5 Attested | arc-challenge | 80.0 ±29 | ★ self | — |
| claude-sonnet-4.5 Attested | commonsenseqa | 60.0 ±33 | ★ self | — |
| claude-sonnet-4.5 Attested | gsm8k | 100.0 ±28 | ★ self | — |
| claude-sonnet-4.5 Attested | mmlu | 62.5 ±28 | ★ self | — |
| claude-sonnet-4.5 Attested | mmlu-pro | 25.0 ±26 | ★ self | — |
| claude-sonnet-4.5 Attested | truthfulqa | 87.5 ±22 | ★ self | — |
Provider Verified · $499/mo
Are you Anthropic API? Subscribe and own this page.
Unlimited multi-model attestations across your hosted catalog, drift alerts to Slack/webhook, customer-facing badge widget, and this anthropic-api page populated daily. See the full pitch →