claude-sonnet-4-5-20250929| Provider | Input, $ per 1M | Output, $ per 1M | In our data since |
|---|---|---|---|
| qihang-ai | $0.43 | $2.14 | 31 Jul 2026 |
| Jiekou.AI | $2.7 | $13.5 | 31 Jul 2026 |
| NanoGPT | $2.992 | $14.994 | 31 Jul 2026 |
| 302.AI | $3 | $15 | 31 Jul 2026 |
| Abacus.AI | $3 | $15 | 31 Jul 2026 |
Amazon Bedrock
anthropic.claude-sonnet-4-5-20250929-v1:0
|
$3 | $15 | 2 Aug 2026 |
| Anthropic | $3 | $15 | 31 Jul 2026 |
| Helicone | $3 | $15 | 31 Jul 2026 |
| LLM Gateway | $3 | $15 | 31 Jul 2026 |
| Merge Gateway | $3 | $15 | 31 Jul 2026 |
Amazon Bedrock
eu.anthropic.claude-sonnet-4-5-20250929-v1:0
|
$3.3 | $16.5 | 2 Aug 2026 |
Amazon Bedrock
bedrock/us-gov-east-1/anthropic.claude-sonnet-4-5-20250929-v1:0
|
$3.6 | $18 | 2 Aug 2026 |
Only the base tier and only per-token prices are shown. Batch, discounted and cached rates, as well as prices per image or per second of video, do not go into this table: they cannot stand in the same column as a price per million tokens.
A single provider appears several times if it sells this model under different identifiers and at different prices — for example, at different weight precision. The identifier is shown next to the name. Identical offers that arrived from two sources under different spellings are merged into one row.
The date in the last column is the day this price first entered our collection. The price may well be older: before that day we simply were not recording it. It has not changed since — otherwise a new row with a new date would stand in its place.
| Task set | Result | Run conditions | Measured by |
|---|---|---|---|
| Arena Score, programming | 1,485.37 1,480–1,490 | — | — |
| Arena Score, long queries | 1,476.02 1,471–1,481 | — | — |
| Arena Score, expert questions | 1,472.22 1,464–1,481 | — | — |
| Arena Score, multi-turn dialogue | 1,464.84 1,459–1,471 | — | — |
| Arena Score, hard prompts | 1,461.7 1,458–1,465 | — | — |
| Arena Score, instruction following | 1,458.64 1,454–1,463 | — | — |
| Arena Score in French | 1,457.25 1,441–1,473 | — | — |
| Arena Score in Spanish | 1,455.76 1,442–1,470 | — | — |
| Arena Score in English | 1,451.19 1,447–1,455 | — | — |
| Arena Score, creative writing | 1,440.99 1,435–1,447 | — | — |
| Arena Score, overall | 1,437.78 1,435–1,441 | — | — |
| Arena Score in Russian | 1,435.09 1,428–1,442 | — | — |
| Arena Score, mathematics | 1,421.66 1,413–1,430 | — | — |
| Arena Score, web development | 1,385.13 1,378–1,392 | — | — |
| MATH, difficulty level five | 97.73 % | thinking budget: 32K · with a tuned harness | Epoch evaluations |
| Mock AIME 2024–2025 — olympiad problems | 77.76 % | thinking budget: 59K · with a tuned harness | Epoch evaluations |
| GPQA Diamond — graduate-level questions | 76.43 % | thinking budget: 16K · with a tuned harness | Epoch evaluations |
| SWE-bench Verified — fixing bugs in repositories | 71.28 % | · with a tuned harness | Epoch evaluations |
| ARC-AGI — generalising to unseen patterns | 63.67 % | thinking budget: 32K · with a tuned harness | — |
| OSWorld — working inside an operating system, first version | 62.9 % | · with a tuned harness | OS World Website |
| Cybench — cybersecurity tasks | 60 % | · with a tuned harness | Opus 4.5 System Card self-reported |
| DeepResearch Bench — deep research | 52.6 % | thinking budget: 2K | DeepResearchBench Leaderboard |
| WeirdML — unusual machine learning tasks | 47.71 % | thinking budget: 16K | WeirdML Leaderboard |
| Terminal-Bench — working in the command line | 46.5 % | · with a tuned harness | Terminal-Bench v2 Leaderboard |
| SimpleBench — trick questions | 45.16 % | thinking budget: 12K | SimpleBench Leaderboard |
| GDPval — tasks from real occupations | 42.5 % | — | — |
| FrontierMath, levels 1–3 | 23.86 % | thinking budget: 32K | Epoch evaluations |
| SimpleQA Verified — factual accuracy | 23.6 % | thinking budget: 59K | Epoch evaluations |
| GSO-Bench — code optimisation | 14.71 % | — | GSO Leaderboard |
| ARC-AGI-2 | 13.61 % | thinking budget: 32K | — |
| Humanity’s Last Exam — expert-level questions | 9.37 % | — | — |
| Chess puzzles | 7.41 % | thinking budget: 32K | Epoch evaluations |
| FrontierMath, level 4 — research-grade problems | 2.44 % | thinking budget: 32K | Epoch evaluations |
| Remote Labor Index — jobs from a freelance marketplace | 2.08 % | — | — |
| CritPt — physics problems | 1.14 % | — | — |