Claude Sonnet 5.5
Anthropic's latest large language model in the Sonnet series, featuring a million-token context window that supports processing of long codebases and extended conversations. It serves as the default reasoning model for Claude Code.
Core Facts
Artificial Analysis gave Argon an intelligence index score of 53, on par with GPT-6 Astra.
Sonnet 5.5 runs more than 30% faster than Sonnet 5.
Claude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.
Timeline (last 90 days)
In the comments of the original Model Trace post, someone used a Python random number generator to successfully "identify" that the random numbers had distilled Claude Sonnet 5.5 and GPT-5 4.
Claude Sonnet 5.5 completed a large batch of changes in about 14 minutes and passed 381 tests; DeepSeek V4.1 Flash built a solid first version of the infrastructure at low cost.
Calculated per thousand repair tasks, Claude Haiku 5.5 costs about $0.70, DeepSeek V4.1 Flash 1017 about $1.35, and MiniMax M3.1 Flash about $4.45 by default and about $0.64 at the low setting.
At 10,000 calls per day, cutting 226 redundant tokens (24% of the prompt) saves about $1,650 per year on Sonnet 5.5.
The tests used Claude Sonnet 5.5 and Claude Opus 5.5 as answering models and Claude Haiku 5.5 as judge, producing 40 sets of results in total.
On Sonnet 5.5, of the 20 skills, 12 were clearly helpful, 4 were clearly worse, and 4 had insufficient samples to judge.
grill-with-docs and hyperframes got clearly worse only on Sonnet 5.5.
In three war tests, five frontier closed-source models (Grok 4.7, Gemini 3.8 Flash, GPT-6 Luna, Meta Muse Spark, Claude Sonnet 5.5) did not trigger midnight even once.
Verified security professionals can use the Claude Mythos 5.1, Opus 5.5 and Sonnet 5.5 models.
Under the adaptive thinking mechanism, the response content array may contain both thinking and text blocks, and the text is not guaranteed to come first, so parsing must filter by type.
40 more timeline events
All Facts (20)
Artificial Analysis gave Argon an intelligence index score of 53, on par with GPT-6 Astra.
80%VerifiedSonnet 5.5 runs more than 30% faster than Sonnet 5.
80%VerifiedClaude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.
75%VerifiedOpus 5.5 scored 66 on the Terminal agent benchmark, higher than Sonnet's 55 and Astra's 57.9.
70%VerifiedHaiku 5.5 can work as a sub-agent alongside Claude Opus 5.5 and Sonnet 5.5 to complete programming tasks.
65%VerifiedEvaluating the per-task cost of large models requires considering both the token unit price and the number of tokens consumed to complete the task; the output token consumption of chain-of-thought reasoning models may increase by several times or even dozens of times.
65%Partially VerifiedThe bundled default model of the cline provider was changed from Claude Sonnet 5.5 to Gemini 4 Argon, because the recommendation list now puts it first.
80%Partially VerifiedIn v3.0.71, when no model is selected, the Cline provider defaults to using the first model in the recommended list, Claude Sonnet 5.5.
75%UnverifiedThe default models for multiple providers were switched to Claude Haiku 5.5, including Google Vertex AI (changed from Claude Sonnet 5.5 to Claude Haiku 5.5), Cortecs, DevPass, Eden AI, GitHub Copilot, LLM Gateway, NanoGPT, OpenCode Go, Requesty, and Vivgrid.
70%Partially VerifiedCalculated per thousand repair tasks, Claude Haiku 5.5 costs about $0.70, DeepSeek V4.1 Flash 1017 about $1.35, and MiniMax M3.1 Flash about $4.45 by default and about $0.64 at the low setting.
65%Partially VerifiedThe test covered four vision models: Claude Sonnet 5 (Anthropic), Gemini 3 Flash (Google), Gemini 3.1 Flash-Lite (Google), and GPT-5.4 nano (OpenAI).
65%Partially VerifiedThe overall assessment of GPT-6.1 Sol is that it is a competent model with high token efficiency and competitive pricing, but it is outperformed in dynamic 3D game visual impact by Claude Sonnet 5.5 at the same price point
65%Partially VerifiedSonnet 5.5 is on par with Opus 5.5 in offensive and defensive cybersecurity capabilities, making it the first model in the Sonnet series to adopt equivalent cybersecurity protection mechanisms.
65%Partially VerifiedSonnet 5.5 is the first Sonnet model released with cybersecurity safeguards, where high-risk cybersecurity tasks will explicitly fall back to Sonnet 5
65%UnverifiedOn Terminal Bench, mid-tier Sonnet 5.5 scores 28.8% at approximately $0.83 per task, while Sonnet 5's highest tier only reaches 10.3% at approximately $11.6 per task
60%UnverifiedClaude Sonnet 5.5 completed a large batch of changes in about 14 minutes and passed 381 tests; DeepSeek V4.1 Flash built a solid first version of the infrastructure at low cost.
50%UnverifiedAt 10,000 calls per day, cutting 226 redundant tokens (24% of the prompt) saves about $1,650 per year on Sonnet 5.5.
50%Unverifiedgrill-with-docs and hyperframes got clearly worse only on Sonnet 5.5.
50%UnverifiedOn Sonnet 5.5, of the 20 skills, 12 were clearly helpful, 4 were clearly worse, and 4 had insufficient samples to judge.
50%UnverifiedThe tests used Claude Sonnet 5.5 and Claude Opus 5.5 as answering models and Claude Haiku 5.5 as judge, producing 40 sets of results in total.
50%