LLM Scorebook
MODEL REFERENCES

Know the model.
Keep the context.

Durable model profiles connect specifications, source evidence, cost and practical guides.

Latest updates
Model profile

Claude Haiku 5.5: a model profile for work you can verify

A practical Claude Haiku 5.5 model profile: specifications, effort settings, benchmark evidence, workload fit, limitations, and an adoption framework.

17 min read · Source-linked editorial

Configurations in the benchmark snapshot

The entries below are sourced configurations, not fourteen finished model profiles.

ConfigurationProviderWeights classificationEvidence
Claude Opus 5.5 (max with fallback)AnthropicClosed weights in source snapshotOriginal source
Claude Sonnet 5.5 (max with fallback)AnthropicClosed weights in source snapshotOriginal source
GPT-6 Astra (max)OpenAIClosed weights in source snapshotOriginal source
Gemini 4 Argon (high)GoogleClosed weights in source snapshotOriginal source
GPT-6 Astra (xhigh)OpenAIClosed weights in source snapshotOriginal source
Muse Spark 1.3 (max)MetaClosed weights in source snapshotOriginal source
Grok 4.7 (xhigh)SpaceXAIClosed weights in source snapshotOriginal source
MiMo-V2.6-ProXiaomiOpen weights; verify licenseOriginal source
Qwen3.8 Max (0902)AlibabaClosed weights in source snapshotOriginal source
GLM-5.3 (max)Z AIOpen weights; verify licenseOriginal source
Step 5 PreviewStepFunClosed weights in source snapshotOriginal source
Kimi K3 (max)KimiOpen weights; verify licenseOriginal source
Ling 3.1 FlashInclusionAIClosed weights in source snapshotOriginal source
Gemini 3.8 Flash (high)GoogleClosed weights in source snapshotOriginal source
Models & reading, in one place. Enter to open
Explore Scorebook
DiscoverLeaderboardCompare modelsLatest updatesThe journalPractical guidesMethodology