Anthropic

Claude Sonnet 3.5

Type
language model
Context
200K tokens
Max output
8K
Released
5 June 2024
Knowledge cutoff
April 2024
API string
anthropic/claude-sonnet-3.5

The provider has deprecated this model. It still responds, but it is not a good choice for new projects.

Prices by provider

Provider Input, $ per 1M Output, $ per 1M In our data since
Poe $2.6 $13 31 Jul 2026
gradient_ai $3 $15 1 Aug 2026
vercel_ai_gateway $3 $15 1 Aug 2026
Cloudflare AI Gateway $3 $15 1 Aug 2026
DigitalOcean Gradient AI $3 $15 1 Aug 2026
Google Vertex AI $3 $15 1 Aug 2026
OpenRouter $3 $15 1 Aug 2026
SAP AI Core $3 $15 1 Aug 2026
Snowflake $3 $15 1 Aug 2026
Replicate $3.75 $18.75 1 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.

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.

Measurement results

Task set Result Run conditions Measured by
Arena Score, programming 1,306.99 1,300–1,314
Arena Score, mathematics 1,303.37 1,296–1,311
Arena Score, multi-turn dialogue 1,299.25 1,292–1,306
Arena Score in French 1,292.38 1,269–1,316
Arena Score in Spanish 1,288.78 1,267–1,311
Arena Score in English 1,287.47 1,283–1,292
Arena Score in Russian 1,281.55 1,273–1,290
Arena Score, overall 1,281.08 1,278–1,284
Arena Score, hard prompts 1,275.69 1,270–1,281
Arena Score, long queries 1,275.23 1,268–1,283
Arena Score, instruction following 1,270.05 1,265–1,275
Arena Score, expert questions 1,246.43 1,235–1,258
Arena Score, creative writing 1,238.49 1,231–1,246
Arena Score, working with images 1,120.44 1,110–1,130
MATH, difficulty level five 51.68 % · with a tuned harness Epoch evaluations
GPQA Diamond — graduate-level questions 38.72 % · with a tuned harness Epoch evaluations
WeirdML — unusual machine learning tasks 30.97 % WeirdML Leaderboard
Cybench — cybersecurity tasks 17.5 % · with a tuned harness Cybench leaderboard
SimpleBench — trick questions 13 % SimpleBench Leaderboard
Mock AIME 2024–2025 — olympiad problems 6.43 % · with a tuned harness Epoch evaluations