
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
Compare Haiku 5.5 Benchmark figures side by side — launch task scores, price tiers and context caps, each tied to its source and retrieval date.
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.

Seedance 2.0
The Future of AI Video Is Here.

Gemini Omni
Gemini Omni Video Generator

AI Multi-Scene Shorts Generator
Create viral AI Shorts instantly

3D Science Video
Create 3D science videos easily

AI Educational Video Generator
Generate Educational videos easily
What a Careful Look at the Haiku 5.5 Benchmark Reveals
Launch-day figures, pricing tiers and per-run costs, each tagged with the configuration it came from and the date it was pulled.
- Where the Launch Figures Come FromElo, partial-credit, coding and accuracy metrics sit on different scales, so averaging them across tables tells you nothing.
- Economics of a Small Model at VolumeSuited to extraction, classification and summarisation jobs that run thousands of times a day.
- Two Footnotes That Rewrite the TablesThe lowest rates end at 100K tokens, and the 72.4% computer-use number is partial credit rather than a completed run.
Turning Haiku 5.5 Benchmark Data into a Decision
A three-step routine for translating published figures into a call about your own workload.
Haiku 5.5 Benchmark Findings at a Glance
Specifications, pricing tiers, coding outcomes, computer-use ratings and third-party index data for Claude Haiku 5.5, all drawn from public sources.
Context Window and Output Ceiling
Text and images go in, text comes out, with adaptive thinking enabled. A one-million-token context and a 128,000-token output cap both tower over Haiku 4.5's 200,000 and 64,000.
Pricing Split by Prompt Length
Pass 100,000 prompt tokens and input, output and caching each shift to a tier priced five times higher. OpenAI's base Standard tier, by contrast, extends to 272,000 input tokens.
Terminal-Bench 4.0: Finished Task Runs
The 39.2% launch result sits clearly above 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: Mergeability and Cost per Effort
Scoring spans correctness, tests, scope, style and repository conventions. At max effort the model reaches 46.4% and at medium 41.6%, at $0.13 per rollout, while stepping from xhigh to max doubles the bill 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. If your workflow has to complete, the 37.1% strict figure is the one to trust.
Third-Party Index Scores and Tool Gains
Artificial Analysis lists the model at all five effort levels on Intelligence Index v4.3.2, with task costs rounded. HLE moves from 45.9% without tools to 57.4% with them, a lift that comes from the surrounding system.
Questions About the Haiku 5.5 Benchmark
Answers on Claude Haiku 5.5 results, token pricing, context limits and how the numbers should be read.
Where does Haiku 5.5 sit next to GPT-6 Luna and GPT-6 Sol?
On knowledge tasks, screenshot work and narrow coding the two are closely matched. Sol and Sol 6.1 charge more and justify it on coding, while Sonnet 5.5 still tops most rows.
Why doesn't the 72.4% computer-use figure count as a success rate?
Require every checkpoint and the same run drops to 37.1%. An agent may find and edit the right file, then save it in the wrong format or quit before the final step. Always add an end-state check.
At what point does Haiku 5.5 move into the higher price tier?
Prompts of 100,000 tokens or less get the lowest rates; beyond that every category is five times dearer. The 90%, 50% and roughly 75% savings figures each measure against a different baseline.
Could Haiku 5.5 run on local hardware, and how much VRAM would that take?
The published specs list no parameter count, active-parameter number, layer count or training compute. 'Small' describes a product tier rather than an architecture, so any VRAM estimate would be guesswork.
Does raising reasoning effort always deliver better results?
No. Sol 6.1 actually scores higher at medium than at max, and a 0.2-point lead over Sonnet at max is far too thin to count as a win. More effort also means more tokens, more latency and more chances to wander off task.
Why does one model post different numbers in different tables?
Sonnet shows 70.6% in Anthropic's launch table but 61.8% on the public leaderboard snapshot, where no Haiku row appears. Artificial Analysis removed its 'Default Fallback' tag, yet the migration guide still states that no server-side fallback exists.
Put the Haiku 5.5 Benchmark to Work on Your Own Tasks
Choose a single workload, fix the effort setting, and see whether the model clears the quality bar you actually need.
