Model Cadence
Plus $4/mo

Claude Haiku 5.5

Anthropic · Released Oct 7, 2026 · 18 days ago · Closed weights
Latest Claude Haiku

Fast Claude model for responsive assistance, classification, and lightweight agents

Strongest at
Computer use
Top 12% of the past year's models · one benchmark so far
Price per 1M tokens
$0.1 / $0.5
input / output · Anthropic
Context window
1M
tokens in · 128K out
Since the last one
357 days
after Claude Haiku 4.5
Lab announcement ↗Open in the timeline

Capability profile

Scored against 90 models released in the past year. 50 is the average model; further out is better.

This modelClaude Haiku 4.5Median of the past year
50100Median of the past yearCoding: 60 · Top 27%Writing: 46 · Better than 23%Reading & documents: 39 · Better than 24%Reasoning: 64 · Top 36%Math: 66 · Top 19%Science: 65 · Top 39%Visual reasoning: 59 · Top 40%Computer use: 86 · Top 12%Claude Haiku 4.5: 57Claude Haiku 4.5: 38Claude Haiku 4.5: 55Claude Haiku 4.5: 52Claude Haiku 4.5: 26Claude Haiku 4.5: 31CodingWritingReadingReasoningMathScienceVisionComputer use
Coding60Top 27%
Writing46−10.9Better than 23% · 1 test
Reading & documents39Better than 24%
Reasoning64+25.7Top 36%
Math66+11.4Top 19%
Science65+13.3Top 39%
Visual reasoning59+32.6Top 40%
Computer use86+54.6Top 12% · 1 test
Where these scores come from

Each capability combines the published benchmarks listed here, rescaled so 50 is the reference average and 15 points is one standard deviation. A hollow dot rests on a single benchmark.

Price against capability

The shaded corner holds anything better and cheaper. Prices blend input and output 3 to 1.

Nothing from the past year scores higher on Computer use for less.

$0.1$0.3$1$3$10406080100Muse Spark Open: 61, $2 per 1MGemini 4 Argon: 109, $4 per 1MGPT-6.1 Sol: 82, $4 per 1MClaude Sonnet 5.5: 92, $4 per 1MGPT-6 Sol: 76, $4 per 1MGPT-6 Luna: 57, $0.2 per 1MClaude Opus 5.5: 93, $8 per 1MGrok 4.7: 70, $3 per 1MDeepSeek V4.1 Flash: 68, $0.27 per 1MGPT-6 Astra: 87, $20 per 1MQwen3.8 Max 0902: 47, $3 per 1MMuse Spark 1.3: 62, $2 per 1MGemini 3.8 Flash: 69, $1.5 per 1MClaude Fable 5.1: 79, $20 per 1MGLM-5.3-Flash: 68, $0.24 per 1MQwen3.8 27B: 77, $1.12 per 1MGLM-5.3: 73, $2.15 per 1MGemini 3.7 Flash: 68, $1.5 per 1MDeepSeek V4 Pro 0813: 76, $0.99 per 1MGrok 4.6: 68, $3 per 1MMuse Spark 1.2: 74, $2 per 1MQwen3.8 Max: 94, $3 per 1MClaude Opus 5: 79, $10 per 1MGemini 3.6 Flash: 66, $1.5 per 1MGemini 3.5 Flash Lite: 47, $0.85 per 1MKimi K3: 70, $6 per 1MGPT-5.6 Terra: 65, $4.5 per 1MGPT-5.6 Sol: 64, $8 per 1MGPT-5.6 Luna: 55, $0.45 per 1MGrok 4.5: 60, $3 per 1MClaude Sonnet 5: 55, $4 per 1MGLM-5.2: 58, $2.15 per 1MKimi K2.7 Code: 51, $1.71 per 1MClaude Fable 5: 78, $20 per 1MClaude Opus 4.8: 59, $10 per 1MGemini 3.5 Flash: 53, $3.38 per 1MDeepSeek V4 Flash: 57, $0.11 per 1MGPT-5.5: 70, $11.2 per 1MKimi K2.6: 34, $1.71 per 1MClaude Opus 4.7: 64, $10 per 1MMuse Spark 1.1: 60, $2 per 1MGLM-5.1: 55, $2.15 per 1MGPT-5.4: 65, $5.62 per 1MGemini 3.1 Pro Preview: 63, $4.5 per 1MClaude Sonnet 4.6: 55, $6 per 1MQwen3.5 397B-A17B: 52, $1.35 per 1MClaude Opus 4.6: 75, $10 per 1MClaude Haiku 4.5: 31, $2 per 1MClaude Haiku 5.5: 86, $0.2 per 1MClaude Haiku 5.5Price per 1M tokens →↑ Computer use$0.1$1$10406080100Muse Spark Open: 61, $2 per 1MGemini 4 Argon: 109, $4 per 1MGPT-6.1 Sol: 82, $4 per 1MClaude Sonnet 5.5: 92, $4 per 1MGPT-6 Sol: 76, $4 per 1MGPT-6 Luna: 57, $0.2 per 1MClaude Opus 5.5: 93, $8 per 1MGrok 4.7: 70, $3 per 1MDeepSeek V4.1 Flash: 68, $0.27 per 1MGPT-6 Astra: 87, $20 per 1MQwen3.8 Max 0902: 47, $3 per 1MMuse Spark 1.3: 62, $2 per 1MGemini 3.8 Flash: 69, $1.5 per 1MClaude Fable 5.1: 79, $20 per 1MGLM-5.3-Flash: 68, $0.24 per 1MQwen3.8 27B: 77, $1.12 per 1MGLM-5.3: 73, $2.15 per 1MGemini 3.7 Flash: 68, $1.5 per 1MDeepSeek V4 Pro 0813: 76, $0.99 per 1MGrok 4.6: 68, $3 per 1MMuse Spark 1.2: 74, $2 per 1MQwen3.8 Max: 94, $3 per 1MClaude Opus 5: 79, $10 per 1MGemini 3.6 Flash: 66, $1.5 per 1MGemini 3.5 Flash Lite: 47, $0.85 per 1MKimi K3: 70, $6 per 1MGPT-5.6 Terra: 65, $4.5 per 1MGPT-5.6 Sol: 64, $8 per 1MGPT-5.6 Luna: 55, $0.45 per 1MGrok 4.5: 60, $3 per 1MClaude Sonnet 5: 55, $4 per 1MGLM-5.2: 58, $2.15 per 1MKimi K2.7 Code: 51, $1.71 per 1MClaude Fable 5: 78, $20 per 1MClaude Opus 4.8: 59, $10 per 1MGemini 3.5 Flash: 53, $3.38 per 1MDeepSeek V4 Flash: 57, $0.11 per 1MGPT-5.5: 70, $11.2 per 1MKimi K2.6: 34, $1.71 per 1MClaude Opus 4.7: 64, $10 per 1MMuse Spark 1.1: 60, $2 per 1MGLM-5.1: 55, $2.15 per 1MGPT-5.4: 65, $5.62 per 1MGemini 3.1 Pro Preview: 63, $4.5 per 1MClaude Sonnet 4.6: 55, $6 per 1MQwen3.5 397B-A17B: 52, $1.35 per 1MClaude Opus 4.6: 75, $10 per 1MClaude Haiku 4.5: 31, $2 per 1MClaude Haiku 5.5: 86, $0.2 per 1MClaude Haiku 5.5Price per 1M tokens →↑ Computer use

Anthropic $0.1 / $0.5 · OpenRouter $0.1 / $0.5 · Cached input $0.01

Since Claude Haiku 4.5

4 releases in the Claude Haiku line.

Claude Haiku 3Mar 13, 2024223dClaude Haiku 3.5Oct 22, 2024358dClaude Haiku 4.5Oct 15, 2025357dClaude Haiku 5.5Oct 7, 2026
Released357 days later
Price in / out$0.1 / $0.5−90%
Context1M5.0×
Writing46−10.9
Reasoning64+25.7
Math66+11.4
Science65+13.3
Visual reasoning59+32.6
Computer use86+54.6
ECI STANDING#16 of 270
SCORE159.32156.19–162.47 CI
FRONTIER LAGLaunched 7.1 months behindScores about what GPT-5.4 Pro did in March 2026.
SCORES LIKE THESE MODELSGPT-5.6 TerraOpenAI · Jul 2026160.09GPT-5.5OpenAI · Apr 2026159.68GPT-5.4 ProOpenAI · Mar 2026159.57Claude Opus 4.8Anthropic · May 2026159.01
Full Epoch Capabilities Index ranking ↗

Epoch benchmark boards

Best result on each board it has run. The tick marks the leader.

OTIS Mock AIME 2024-2025
99%#19 of 215
ProofBench
89%#9 of 81
DTBench
95%#23 of 179
Chess Puzzles
55%#16 of 93
GPQA diamond
90%#46 of 233
FrontierMath-Tiers-1-3-v2-Private
75%#20 of 89
ARC-AGI-2
61%#30 of 103
ARC-AGI
89%#32 of 107
9 more boards
FrontierMath-Tier-4-v2-Private
46%#23 of 71
FrontierCode
46%#15 of 46
Mystery Game Puzzles
30%#29 of 87
APEX-Agents
56%#17 of 51
EBR-bench
44%#9 of 27
Furniture Assembly
48%#12 of 34
FrontierSWE
44%#8 of 21
WeirdML v3
24%#11 of 23
SimpleQA Verified
24%#68 of 84

Anthropic's past year

12 release days in the past year, about one every 30 days. The next is expected around Nov 17, 2026. Forecasts ↗

NovDecJanFebMarAprMayJunJulAugSepOctNov

Specifications

RELEASE DATEOct 7, 2026
WEIGHTSClosed weights
CONTEXT WINDOW1M
MAX OUTPUT128K
REASONINGNot listed
TOOL CALLINGNot listed
INPUTStext, image/jpeg, image/png, image/gif, image/webp, application/pdf
OUTPUTStext

Models.dev record ↗

Specifications and prices from Models.dev; availability and limits may differ by provider. Benchmarks from Epoch AI (CC BY 4.0), LiveBench and the other sources named above.