Pass IndexThe State of AISign in

Claude 3.7 Sonnet

Claude 3.7 Sonnet was Anthropic's flagship model at release and the first hybrid reasoning model on the market, combining fast standard responses with an optional extended thinking mode.

text + image → text · made by Anthropic

$3→$15per Mtok in / out
AWS Bedrock · 5 sellers

Sold by 6 ways

SellerLaneRate
Databricks 0 ways$3per Mtok in$15per Mtok out$0.3per Mtok cached$3.75per Mtok cache write
Databricks cloudnot on the seller's list todaystandard$3per Mtok in$15per Mtok out$0.3per Mtok cached$3.75per Mtok cache writeraw.githubusercontent.com · read 2026-09-22
AWS Bedrock cloudstandard$3per Mtok in$15per Mtok out$0.3per Mtok cached$3.75per Mtok cache writeraw.githubusercontent.com · read 2026-10-31
DigitalOcean cloudstandard$3per Mtok in$15per Mtok out$0.3per Mtok cached$3.75per Mtok cache writeraw.githubusercontent.com · read 2026-10-31
Snowflake Cortex cloudstandard$3per Mtok in$15per Mtok out$0.3per Mtok cachedraw.githubusercontent.com · read 2026-10-31
Replicate aggregatorstandard$3per Mtok in$15per Mtok outreplicate.com · read 2026-08-25
DeepInfra aggregatorstandard$3.3per Mtok in$16.5per Mtok outapi.deepinfra.com · read 2026-08-25

Measured 27 standings

PlaceBoardMetricScore
3rdof 54Live bench — Epoch AIGlobal average76.1
4thof 16The agent company — Epoch AI% Score0.402
5thof 15Cad eval — Epoch AIOverall pass (%)0.54
5thof 8Spatialviz bench — Epoch AIOverall score0.339
9thof 20Os world — Epoch AIScore35.8
12thof 22Cybench — Epoch AIUnguided % Solved0.2
17thof 49Lech mazur writing — Epoch AIMean score8.11
19thof 69Aider polyglot coding leaderboardPercent correct64.9
20thof 108Math level 5 — Epoch AImean_score91.163
21stof 38Vpct — Epoch AICorrect0.39
23rdof 72Aider polyglot — Epoch AIPercent correct64.9
23rdof 42Fictionlivebench — Epoch AI120k token score0.531
23rdof 49Metr time horizons — Epoch AIaverage_score0.6
27thof 46Enigma eval — Epoch AIAccuracy4.23
31stof 35Swe bench verified — Epoch AImean_score60.95
32ndof 41Deepresearchbench — Epoch AIAverage score0.436
36thof 42Gso — Epoch AIScore OPT@10.038
39thof 92Forecastbench — Epoch AIOverall score64
39thof 56Humanity's Last Exam (Epoch AI replication)Accuracy8.04
59thof 102Simplebench — Epoch AIScore (AVG@5)0.464
75thof 101FrontierMath (Tiers 1-3)mean_score4.138
105thof 165LMArena · VisionScore (Elo)1175
143rdof 326GPQA Diamondmean_score78.504
182ndof 306OTIS Mock AIME 2024-2025mean_score57.778
195thof 414LMArena · TextScore (Elo)1372
249thof 309Arc agi — Epoch AIScore28.6
257thof 290Arc agi 2 — Epoch AIScore0.9

About

Claude 3.7 Sonnet — a text model from Anthropic, sold by 5 companies from $3 in and $15 out per million tokens, placed 3rd of 54 on Live bench — Epoch AI.

It takes text and images and returns text, with a context window of 200,000 tokens. It was published in February 2025, trained on material up to October 2024. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning, code and agents. Five companies sell it. The cheapest is $3 in and $15 out per million tokens at AWS Bedrock. It has been measured on 27 boards, and stands best at 3rd of 54 on Live bench — Epoch AI.

Every current figure

Maker
Anthropic
Register
model
Takes
text + image
Returns
text
Context
200,000 tokens
Longest answer
64,000 tokens
Published
February 2025
Knowledge to
October 2024
Licence
not read
Sellers
5
Maker's own price
not read
Price
$3 in and $15 out per million tokens — AWS Bedrock
Boards
27
Best place
3rd of 54 — Live bench — Epoch AI

Known as 8 names

Claude Sonnet 3.7 (2025-02-19)Claude Sonnet 3.7 Thinking (2025-02-19)claude-3-7-sonnet-20250219 (32k thinking tokens)claude-3-7-sonnet-20250219 (no thinking)OpenHands-Versa (Claude-3.7 Sonnet)Claude-Sonnet-3.7anthropic/claude-3-7-sonnet-latestanthropic/claude-3.7-sonnet