Claude Sonnet 4.6
A model in Anthropic's Claude Sonnet series, previously rumored to be Sonnet 5 under the codename Phoenix, but officially released as version 4.6.
Timeline (last 90 days)
The video's author used Model Trace to test DeepSeek V4.1 Flash 1070, and the result was identified as Claude Sonnet 4.6 with a probability of 99.9%.
Claude Sonnet 4.6 prompt caching pricing: cache writes (5 minutes) $3.75/MTok, cache writes (1 hour) $6.00/MTok, cache reads $0.30/MTok.
In head-to-head user preference comparisons among early testers, 59% preferred Claude Sonnet 4.6 over Claude Opus 4.5.
The Batch API's 50% discount stacks with long context, and the combined long context plus Batch price is $3/MTok input and $11.25/MTok output.
The web search tool is charged at $10 per thousand searches, with each search call counting as one, and failed searches are not billed.
When the code execution tool is used on its own, each organization gets 1,550 free hours per month, with usage beyond that charged at $0.05 per container-hour, billed in minimum increments of 5 minutes.
In computer use, the system prompt adds 466–499 tokens, each tool definition adds 735 tokens, and each screenshot consumes roughly 2,000–5,000 tokens.
Claude Opus 4.6 has a maximum output of 128K tokens, and Claude Sonnet 4.6 has a maximum output of 64K tokens.
The cache price for Anthropic Sonnet 4.6 is listed as "at input rate," meaning this model does not offer a separate cache discount tier.
Manual Extended Thinking (thinking.type: "enabled" plus budget_tokens) is already marked as outdated on Opus 4.6 and Sonnet 4.6, is not accepted at all by the new models, and is now controlled by the effort parameter, with a default of medium.
4 more timeline events
All Facts (13)
When the code execution tool is used on its own, each organization gets 1,550 free hours per month, with usage beyond that charged at $0.05 per container-hour, billed in minimum increments of 5 minutes.
50%UnverifiedIn computer use, the system prompt adds 466–499 tokens, each tool definition adds 735 tokens, and each screenshot consumes roughly 2,000–5,000 tokens.
50%UnverifiedClaude Opus 4.6 has a maximum output of 128K tokens, and Claude Sonnet 4.6 has a maximum output of 64K tokens.
50%UnverifiedThe Batch API's 50% discount stacks with long context, and the combined long context plus Batch price is $3/MTok input and $11.25/MTok output.
50%UnverifiedThe web search tool is charged at $10 per thousand searches, with each search call counting as one, and failed searches are not billed.
50%UnverifiedIn head-to-head user preference comparisons among early testers, 59% preferred Claude Sonnet 4.6 over Claude Opus 4.5.
50%UnverifiedClaude Sonnet 4.6 prompt caching pricing: cache writes (5 minutes) $3.75/MTok, cache writes (1 hour) $6.00/MTok, cache reads $0.30/MTok.
50%UnverifiedThe cache price for Anthropic Sonnet 4.6 is listed as "at input rate," meaning this model does not offer a separate cache discount tier.
50%UnverifiedManual Extended Thinking (thinking.type: "enabled" plus budget_tokens) is already marked as outdated on Opus 4.6 and Sonnet 4.6, is not accepted at all by the new models, and is now controlled by the effort parameter, with a default of medium.
50%UnverifiedAll models failed on EMI compound interest calculations; for case e04 (₹72,00,000 loan, 8.4% annual interest rate, 25-year term) the correct answer is ₹57,492, Claude Sonnet 4.6 answered ₹57,487, and Gemini 3.1 Pro Preview answered ₹57,482.
50%UnverifiedEven in the Indian context, 5% to 12% of responses from all models used Western thousands-separator formatting (Western comma drift) in intermediate steps, with Gemini 3.7 Flash the lowest (5%) and Claude Sonnet 4.6 the highest (12%).
50%UnverifiedClaude Sonnet 4.6 answered ₹19 lakh for case c14 (find the 20% down payment in rupees for a ₹95 lakh house), which was mathematically correct but had the wrong unit (the question required rupees).
50%UnverifiedThe video's author used Model Trace to test DeepSeek V4.1 Flash 1070, and the result was identified as Claude Sonnet 4.6 with a probability of 99.9%.
50%