Anthropic, released May 22, 2025

Claude Sonnet 4price, context, benchmarks and release details

Provisional: not enough results to rank yet 100% confidence 100 percent, Full confidence, 7 of 10 expected sources in
46.6
SI Score
#115 of 142 ranked models
Input, per 1M tokens
$3.00Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Output, per 1M tokens
$15.00Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Context window
200Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Max output
64Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released
May 22, 2025models.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗

How the score breaks down

Coding (weight 40 percent) 45.1
Math (weight 15 percent) 74.8
Preference (weight 15 percent) 66.7
Reasoning (weight 30 percent) 21.9

Weights: reasoning 30%, math 15%, coding 40%, preference 15%. Results use fixed 0–100 scales before averaging, and thin evidence is pulled toward 50. Method si-v3-retained-evidence-2, computed Oct 9, 2026, 06:15 UTC.

Around it on the leaderboard

  1. 113 DeepSeek-R1 47.9
  2. 114 GLM-4.6 47.8
  3. 115 Claude Sonnet 4 46.6
  4. 116 Nemotron 3.5 Lightning 30B A3B 46.1
  5. 117 GPT-5 Pro 45.8

Full leaderboard

Benchmark results

12 benchmarks, 39 results

Each row shows the best published result. Where a model was tested at several settings, such as reasoning effort, open the row to see each one. Hover or tap a value for its source.

ARC-AGI-1 (public eval)reasoning 56.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 56.8 4 settings
  • Setting 1 33%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514Published Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 2 56.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 3 31.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-1kPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 4 48.6%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-8k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-1 (public eval)
ARC-AGI-1 (semi-private)reasoning 40%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 40.0 4 settings
  • Setting 1 23.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514Published Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 2 40%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 3 28.0%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-1kPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 4 29.0%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-8k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-1 (semi-private)
ARC-AGI-2 (public eval)reasoning 4.4%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 4.4 4 settings
  • Setting 1 2.1%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514Published Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 2 4.4%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 3 1.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-1kPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 4 2.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-8k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-2 (public eval)
ARC-AGI-2 (semi-private)reasoning 5.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 5.9 4 settings
  • Setting 1 1.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514Published Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 2 5.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-16k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 3 0.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-1kPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 4 2.1%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. [variant] claude-sonnet-4-20250514-thinking-8k-bedrockPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-2 (semi-private)
GPQA Diamondreasoning 78.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 32KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 78.3 4 settings
  • 16K thinking 75.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 16KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • 32K thinking 78.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 32KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • 59K thinking 77.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 59KPublished May 26, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 4 66.7%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About GPQA Diamond
Humanity's Last Exam (Scale AI)reasoning 5.5%Humanity’s Last ExamPotential contamination warning: This model was evaluated after the public release of HLE, allowing model builder access to the prompts and solutions. [variant] Published May 23, 2025 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 5.5
MATH Level 5math 84.4%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 84.4
OTIS Mock AIME 2024–2025math 71.1%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 32KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 71.1 4 settings
  • 16K thinking 53.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 16KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • 32K thinking 71.1%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 32KPublished May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • 59K thinking 68.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] 59KPublished May 23, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 4 28.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published May 22, 2025 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About OTIS Mock AIME 2024–2025
Aider Polyglotcoding 56.4%Aider polyglotPublished source fact [variant] Aider polyglot; 225 cases; 2 attemptsPublished May 24, 2025 Retrieved Oct 9, 2026 · Apache-2.0
Open source ↗ 56.4
SWE-bench Pro (public)coding 9.1%SWE-bench Pro (public)Published steward score [variant] Published Sep 19, 2025 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 9.1
SWE-bench Verifiedcoding 76.8%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] EPAM AI/Run Developer AgentPublished Aug 4, 2025 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 76.8 10 settings
  • Setting 1 57%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] Artemis Agent v2Published Sep 24, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 2 71.2%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] BloopPublished Jul 10, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 3 76.8%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] EPAM AI/Run Developer AgentPublished Aug 4, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 4 74.8%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] Harness AIPublished Jul 31, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 5 74.6%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] Lingxi-v1.5Published Jul 20, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 6 64.9%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] mini-SWE-agent; 1.0.0Published Jul 26, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 7 70.8%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] Moatless ToolsPublished Jun 11, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 8 70.4%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] OpenHandsPublished May 24, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 9 66.6%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] SWE-agentPublished May 22, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • Setting 10 72.4%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] ToolsPublished May 22, 2025 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About SWE-bench Verified
LMArena Textpreference 1339.3 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 8, 2026 Retrieved Oct 9, 2026 · CC-BY-4.0
Open source ↗ 66.7

Normalization uses a fixed 0–100 scale for each unit, independent of other models. Compare evaluation conditions before reading a small gap as decisive. “Lab-reported” marks the provider's own published figure.

Details and sources

Open weights
Nomodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
License
not yet reported
Input modalities
text, image, pdfmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
First seen by SuperIndex
Oct 8, 2026
Coverage
82% of expected source weight

Reported (7)

  • Aider polyglotOct 8, 2026
  • ARC PrizeOct 8, 2026
  • Epoch AI BenchmarkingOct 8, 2026
  • Humanity’s Last ExamOct 8, 2026
  • LMArena / ArenaOct 8, 2026
  • SWE-bench VerifiedOct 8, 2026
  • SWE-bench Pro (public)Oct 8, 2026

Awaiting (3)

  • Official model cards via models.dev3% of weight
  • LiveBench10% of weight
  • Terminal-Bench5% of weight

Confidence rises as pending sources publish. Some sources never cover some models, so confidence reaches 100% at 80% of expected weight.

What changed

All releases · RSS feed

Alerts on this device

What to be alerted about
RSS feed