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 results side by side: task scores, cost bands and context ceilings, each figure tied to its source.
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
Turning the Haiku 5.5 Benchmark Into a Decision
Launch-day figures, tiered costs and per-effort spend, each stored beside the setup it came from and the day it was pulled.
- What Those Launch Figures Really CaptureElo ratings, partial credit, coding passes and accuracy rates sit on scales that cannot be averaged together.
- Cost Math for Work That RepeatsDesigned for extraction, tagging and summarising jobs that run thousands of times over.
- Two Footnotes That Reshape the TablesThe cheapest rates end at 100K tokens, and the 72.4% computer-use figure counts partial credit, not finished runs.
Putting Claude Haiku 5.5 Results to Work
A three-step path that turns published Haiku 5.5 benchmark figures into a confident call on your own workload.
What Stands Out for Claude Haiku 5.5
Spec sheets, cost bands, coding outcomes, computer-use results and third-party index data behind the Haiku 5.5 Benchmark.
Context Window and Output Ceiling
Text and images go in, text comes out, and adaptive thinking stays on. A one-million-token context plus a 128,000-token output cap make Haiku 4.5's 200,000 and 64,000 look small.
Pricing Split by Prompt Length
Once a prompt passes 100,000 tokens, input, output and caching all move to a tier priced five times higher. By contrast, OpenAI's base Standard tier extends to 272,000 input tokens.
Terminal-Bench 4.0: Tasks Finished
At launch, Haiku's 39.2% clears Luna by a clear margin yet stays below the published Sol entries. Agent runs allow an eight-hour timeout, and any 2.1 or 3.0 figure comes from an earlier experiment.
FrontierCode: Mergeability and Effort Cost
Scoring weighs correctness, tests, scope, style and repository conventions. Haiku reaches 46.4% at max effort and 41.6% at medium for $0.13 per rollout, and stepping from xhigh to max doubles the spend for half a point.
OSWorld 2.1: Partial Credit vs. a Strict Pass
The run is defined by 82 offline tasks, five attempts each, 1080p screenshots and a 500-step limit. If a workflow has to finish, the 37.1% strict figure is the one to trust.
Third-Party Index Scores and the Tool Boost
Artificial Analysis lists Haiku across all five effort settings on Intelligence Index v4.3.2, with task costs rounded. HLE shows 45.9% bare and 57.4% with tools — an uplift that comes from the surrounding setup.
Haiku 5.5 Benchmark — FAQ
Straight answers on Claude Haiku 5.5 evaluation numbers, what each tier costs and where the ceilings fall.
How does Haiku 5.5 stack up against GPT-6 Luna and GPT-6 Sol?
Across knowledge tasks, screenshot handling and tight coding work, Haiku stays level with Luna. Sol and Sol 6.1 ask for more money and justify it on coding, yet Sonnet 5.5 still tops most rows.
Why doesn't the 72.4% computer-use figure count as a success rate?
Under strict grading, where every checkpoint has to pass, the same run lands at just 37.1%. A model may find and edit the correct file, then save the wrong format or halt one action early — so always add an end-state check.
At what point does Haiku 5.5 move into the higher price band?
The lowest rates apply only to prompts of 100,000 tokens or less; past that line, each category is billed at five times the cheap tier. Anthropic's 90%, 50% and roughly 75% savings numbers each measure against a different baseline.
Could Haiku 5.5 run on local hardware, and what GPU memory would that take?
The published specs list 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 rests on nothing.
Does pushing reasoning effort higher always deliver 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 far too thin to count as a win. More effort also means more tokens, more latency and more room to wander off task.
Why does one model post different numbers across different tables?
Sonnet appears at 70.6% in Anthropic's launch table but 61.8% on the public leaderboard snapshot, where Haiku shows no visible row at all. Artificial Analysis has removed its 'Default Fallback' tag, while Anthropic's migration guide still states that no server-side fallback exists.
Run the Haiku 5.5 Benchmark Against Your Own Work
Choose a single job, dial in the effort level and see whether Haiku 5.5 clears the bar you set.