Claude 3.5 Sonnet
A model released by Anthropic in the Claude 3.5 family, achieving a pass rate exceeding 80% on programming benchmarks. Its strong code generation capabilities make it one of the mainstream AI coding tools used in vibe coding workflows.
Core Facts
The context window in the GPT-3 era was only 2,048 tokens, whereas Google Gemini already supports millions of tokens, Claude supports 200K tokens, and GPT-4 Turbo supports 128K tokens.
Claude 3.5 Sonnet has a context window of 200K tokens
Claude 3.5 Sonnet supports a context length of 200K tokens, capable of processing approximately 150,000 Chinese characters at once.
Timeline (last 90 days)
Using GPT-4o or Claude 3.5 Sonnet to handle Agent intermediate steps (formatting, simple classification, tool-call validation) breaks unit economics.
If closed-source vision APIs such as GPT-4o or Claude 3.5 Sonnet were called, at $0.03 to $0.05 per receipt, a restaurant with about 100 receipts per day, or 3,000 per month, would face an API bill of $90 to $150 per month.
METR's 2025 trial mainly used Cursor Pro paired with early tools such as Claude 3.5 and 3.7 Sonnet.
Amazon Web Services (AWS) issued a formal notice that Claude 3 Haiku, Claude 3 Sonnet, Claude 3.5 Sonnet, Claude 3.5 Sonnet v2, and Claude 3.7 Sonnet on Amazon Bedrock will cease service on November 13, 2026 in seven specific regions.
In the VibeCodeArena leaderboard snapshot, Claude 3.5 Sonnet ranks 3rd with an eval score of 87.5, Claude Opus 5 ranks 4th with an eval score of 94.0, and Llama-4-Maverick ranks 9th with an eval score of 91.3.
The routing layer can be divided into three tiers: the local/edge tier (such as Llama 3 8B, Mistral 7B, cost about $0.05–$0.15/1M tokens, latency <250ms), the mid-tier hosted tier (such as Claude 3.5 Haiku, GPT-4o-mini, cost about $0.25–$0.60/1M tokens, latency <800ms), and the frontier reasoning tier (such as Claude 3.5 Sonnet, o1, GPT-4o, cost about $3.00–$15.00/1M tokens, latency 2s–8s).
Anthropic isolates rate limits by model series, so Claude 3.5 Sonnet requests use a different bucket from Haiku or Opus.
Claude 3 Opus, Claude 3.5 Sonnet, and the not-yet-released Opus 5.5 do not share some unified cross-version failure mode theory.
DataBugBench testing covers five models: Claude 3.5 Sonnet, GPT-4o, Gemini 1.5 Pro, Qwen 2.5 Coder 32B, and Llama 3.3 70B.
When generating Rust code, Claude 3.5 Sonnet consistently finds the optimal zero-cost abstractions and passes the borrow checker on the first attempt.
36 more timeline events
All Facts (20)
The context window in the GPT-3 era was only 2,048 tokens, whereas Google Gemini already supports millions of tokens, Claude supports 200K tokens, and GPT-4 Turbo supports 128K tokens.
90%VerifiedClaude 3.5 Sonnet has a context window of 200K tokens
90%VerifiedClaude 3.5 Sonnet supports a context length of 200K tokens, capable of processing approximately 150,000 Chinese characters at once.
80%VerifiedThe Computer Use technology was first publicly demonstrated by Anthropic in October 2024 at the launch of Claude 3.5 Sonnet.
80%VerifiedClaude 3.5 Sonnet is a model released by Anthropic in 2024 that surpassed the contemporaneous GPT-4o on the code generation benchmark HumanEval
80%VerifiedFrontier-grade is a term used in the AI industry to describe the capability tier of the current top large models, typically referring to models such as GPT-4, Claude 3, and Gemini Ultra.
75%VerifiedThe Artifacts feature was first introduced by Anthropic in June 2024 alongside Claude 3.5 Sonnet.
75%VerifiedLocally run open-source models generally struggle to match top-tier cloud models such as GPT-4 and Claude in capability.
70%VerifiedOn mainstream benchmarks such as MMLU and HumanEval, the score gaps among Claude 3.5, GPT-4o, Gemini 1.5 Pro, and Grok-2 have been compressed to within the range of statistical error.
70%VerifiedCode review tasks have relatively fixed patterns and predictable context scope, making this type of highly structured work the easiest to achieve substantial gains through framework optimization and model efficiency improvements.
65%VerifiedClaude 3.5 Sonnet has approached or even surpassed the performance of the early Claude 3 Opus in most benchmark tests.
65%VerifiedClaude 3.5 Sonnet and the Claude 4 series have performed prominently on authoritative benchmarks such as SWE-bench.
65%UnverifiedClaude 3.5 Sonnet在2024年中期发布后因性价比极高而广受欢迎
85%UnverifiedAnthropic's Claude 3.5 Sonnet charges about $3 per million Tokens for input and $15 per million Tokens for output
80%UnverifiedAnthropic此前已在Claude 3.5 Sonnet上展示过在不改变模型权重的情况下通过工程手段实现显著加速的能力
70%Unverified当前主流开放权重模型与Claude 3.5/3.7 Sonnet之间在代码复杂推理、长上下文理解和多步任务执行方面仍存在可感知的能力差距
70%UnverifiedPerplexity combines retrieval-augmented generation (RAG) with a multi-model routing strategy, and the Pro version allows users to choose the underlying model, including GPT-4o, Claude 3.5 Sonnet, Sonar Large, and others.
60%UnverifiedClaude 3.5 Sonnet is Anthropic's first model to support Computer Use.
60%UnverifiedUsing GPT-4o or Claude 3.5 Sonnet to handle Agent intermediate steps (formatting, simple classification, tool-call validation) breaks unit economics.
50%UnverifiedIf closed-source vision APIs such as GPT-4o or Claude 3.5 Sonnet were called, at $0.03 to $0.05 per receipt, a restaurant with about 100 receipts per day, or 3,000 per month, would face an API bill of $90 to $150 per month.
50%