None
None
Long Story Video Skill
Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.
Ads Video Skill
Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.
3D Science Explainer Video Skill
Convert scientific concepts into stunning 3D explain animations
Feedback
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
Haiku 5.5 Benchmark
Task scores, two token price tiers and context caps from the Haiku 5.5 Benchmark — every figure carries a note on where it came from.
All Tools
Discover our comprehensive AI-powered animation toolkit
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
Veo3.1
Create Stunning Videos with Veo3.1
FLUX 3 Video Generator
Kling 3.0
Next-Gen AI Video Generator
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
A Clear-Eyed Look at the Haiku 5.5 Benchmark
Reported scores, two token price tiers and per-effort spend, each paired with the settings and the date it was captured.
- Which Scales the Reported Scores UseElo ratings, partial-credit results, coding passes and accuracy rates sit on incompatible scales, so averaging them across rows misleads.
- Economics of a Small Model at High VolumeDesigned for extraction, classification and summarization jobs that run thousands of times a day.
- Two Footnotes That Reshape the TablesThe lowest rates end at 100K tokens, and 72.4% on computer use reflects partial credit rather than completed tasks.
Three Steps to a Haiku 5.5 Benchmark Decision
A short process for converting reported Haiku 5.5 Benchmark results into a call about your own workload.
What Stands Out in the Haiku 5.5 Benchmark
Reported specs, two pricing tiers, coding outcomes, computer-use results and third-party index data for Claude Haiku 5.5.
Context Size and Output Ceiling
Text and images go in, text comes out, with adaptive thinking enabled. A 1M-token context and a 128,000-token output limit far exceed Haiku 4.5's 200,000 and 64,000.
Pricing Splits by Prompt Length
Past 100,000 prompt tokens, input, output and cache pricing each move to a tier five times the lower one. OpenAI's base Standard tier extends to 272,000 input tokens.
Terminal-Bench 4.0 Task Completion
At launch Haiku posted 39.2%, comfortably ahead of Luna yet below the published Sol entries. Runs allow an eight-hour agent timeout, and any 2.1 or 3.0 figures come from an earlier experiment.
FrontierCode Merge Rate and Cost per Effort
Scoring weighs correctness, tests, scope, style and repo conventions. Max effort yields 46.4% and medium 41.6% at $0.13 per rollout, yet going from xhigh to max doubles cost for half a point.
OSWorld 2.1: Partial Credit vs. Strict Pass
The setup uses 82 offline tasks, five tries each, 1080p screenshots and a 500-step cap. When a workflow has to complete, the 37.1% strict figure is the one that counts.
Third-Party Index Scores and Tool Gains
Artificial Analysis lists Haiku across all five effort settings on Intelligence Index v4.3.2 with rounded task costs. HLE shows 45.9% tool-free and 57.4% with tools, a lift that comes from the surrounding setup.
Haiku 5.5 Benchmark: Questions Answered
Straight answers on Claude Haiku 5.5 benchmark scores, token pricing and hard limits.
How does Haiku 5.5 stack up against GPT-6 Luna and GPT-6 Sol?
Across knowledge tasks, screenshot work and small coding jobs, Haiku and Luna trade wins. Sol and Sol 6.1 charge more and justify it on coding, yet Sonnet 5.5 still tops most categories.
Why doesn't the 72.4% computer-use figure equal a success rate?
Require every checkpoint and the same run drops to 37.1%. An agent may find and edit the correct file, then save it in the wrong format or halt before the final action. Add an end-state check.
At what point does Haiku 5.5 move to the higher price tier?
The lowest rates apply only to prompts of 100,000 tokens or fewer; beyond that every category costs five times as much. Anthropic's 90%, 50% and roughly 75% savings numbers each use a different baseline.
Could Haiku 5.5 run on local hardware, and how much VRAM would that take?
The published specs give no parameter count, active-parameter total, layer count or training compute. 'Small' describes a product tier rather than an architecture, so any local GPU estimate is guesswork.
Does raising reasoning effort always produce better results?
No. Sol 6.1 actually scores higher at medium than at max, and Haiku's 0.2-point lead over Sonnet at max is too narrow to count as a win. More effort means more tokens, more latency and more chances to wander off task.
Why do identical models post different figures across tables?
Sonnet shows 70.6% in Anthropic's launch table but 61.8% in the public leaderboard snapshot, where Haiku has no visible row. Artificial Analysis removed its 'Default Fallback' tag, while Anthropic's migration guide still states that no server-side fallback exists.
Benchmark Haiku 5.5 on Your Own Workload
Choose a single task, fix the effort setting, and see whether Haiku 5.5 clears the quality bar you already use.