Name: Claude Sonnet 4.5 Released
Author: Anthropic

Claude Sonnet 4.5 Released · Chronopin

AIPosted 09/28/202609/28/2026 at 9:03 pm

To timeline

Anthropic

Starts 09/29/2025

San Francisco, California, United States

To map

Thread

Respond to this Pin

SWE-bench Verified%

Gemini 2.5 Pro67.2
Claude Sonnet 472.7
GPT-572.8
Claude Opus 4.174.5
Claude Sonnet 4.577.2

source

Terminal-Bench%

Gemini 2.5 Pro25.3
Claude Sonnet 436.4
GPT-543.8
Claude Opus 4.146.5
Claude Sonnet 4.550

source

OSWorld%

Claude Sonnet 442.2
Claude Opus 4.144.4
Claude Sonnet 4.561.4

source

Finance Agent%

Gemini 2.5 Pro29.4
Claude Sonnet 444.5
GPT-546.9
Claude Opus 4.150.9
Claude Sonnet 4.555.3

source

Claude Sonnet 4.5 specs as launched

Spec	Value
Maker	Anthropic^[1][3]
API model ID	claude^[2]-sonnet-4-5
API price per 1M tokens

Language model rubric2026 criteria

Alternate names

Claude Agent SDK
Alias

User Photos

Photos from readers, kept apart from the pin's sourced media. Rate what you ate or saw, and on a restaurant pin tag the dishes right on your photo.

Comments

References 589% CONFIDENCEOverall confidence: 89%How well the pin's source and references back up its dates.Weighted average of how firmly 5 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better

The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.

[1]90%
anthropic.com/news/claude-sonnet-4-5
anthropic.com· Posted Sep 28, 2026· Starts Sep 29, 2025 ✓· 29% of score
The post is dated "Sep 29, 2025" and says "Claude^[2] Sonnet 4.5 is available everywhere today"; the docs model page lists "Released September 29, 2025" and TechCrunch^[4] reports the launch "On Monday".
[2]90%
Claude Sonnet 4.5 - Claude Platform Docs
platform.claude.com· Added Sep 28, 2026· 29% of score
Anthropic's^[1][3] model page: claude-sonnet-4-5-20250929, 200K context window, 64K max output, $3/$15 per million tokens, extended thinking, text and images in, a January 2025 reliable knowledge cutoff, "Released September 29, 2025".
[3]90%
System Card: Claude Sonnet 4.5
anthropic.com· Added Sep 28, 2026· 29% of score
Anthropic's^[1] system card for Claude^[2] Sonnet 4.5, "a new hybrid reasoning large language model" (September 2025), with the alignment and safety evaluations behind its ASL-3 release.
[4]85%
Anthropic launches Claude Sonnet 4.5, its best AI model for coding
techcrunch.com· Published Sep 29, 2025· Starts Sep 29, 2025· 7% of score
TechCrunch, 2025-09-29: "On Monday, Anthropic^[1][3] launched a new frontier model called Claude^[2] Sonnet 4.5" at Sonnet 4's $3/$15; researcher David Hershey saw it code autonomously for up to 30 hours; it arrives less than two months after Claude Opus 4.1.
[5]85%
Anthropic launches Claude Sonnet 4.5, its latest AI model that's 'more of a colleague'
cnbc.com· Published Sep 29, 2025· Starts Sep 29, 2025· 7% of score
CNBC, 2025-09-29: available to all users from the $183 billion startup; Mike Krieger says it will be the default and recommends it for "basically every use case"; it runs autonomously for 30 hours against Opus 4's seven.

Suggest a correction

Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.

Reviews

Claude Sonnet 4.5 is probably the "best coding model in the world" (at least for now) · simonwillison.net · Sep 29, 2025My initial impressions were that it felt like a better model for code than GPT-5-Codex, which has been my preferred coding model since it launched a few weeks ago
Claude Sonnet 4.5 Is A Very Good Model · thezvi.substack.com · Oct 1, 2025My tentative evaluation is a qualified yes. This is likely a big leap in some ways. If I had to pick one 'best coding model in the world' right now it would be Sonnet 4.5.
Claude Sonnet 4.5: Better performance but a paradox · coderabbit.ai · Oct 3, 2025Sonnet 4.5 gets you close to Opus-level performance at a fraction of the price, making it a pragmatic sweet spot for teams reviewing code at scale.
Rebuilding Devin for Claude Sonnet 4.5: Lessons and Challenges · cognition.ai · Sep 29, 2025With Sonnet 4.5, we're seeing the biggest leap since Sonnet 3.6 (the model that was used with Devin's GA): planning performance is up 18%, end-to-end eval scores up 12%, and multi-hour sessions are dramatically faster and more reliable.
We Tested Claude Sonnet 4.5 for Writing and Editing · every.to · Oct 23, 2025Then Anthropic dropped Sonnet 4.5, and within 48 hours my workflow collapsed from two chat interfaces into one.

Sentiment over timePositive

Sentiment

Sentiment over time
Claude Sonnet 4.5 is probably the "best coding model in the world" (at least for now) · Sep 29, 2025	0.80
Rebuilding Devin for Claude Sonnet 4.5: Lessons and Challenges · Sep 29, 2025	0.80
Claude Sonnet 4.5 Is A Very Good Model · Oct 1, 2025	0.80
Claude Sonnet 4.5: Better performance but a paradox · Oct 3, 2025	0.70
We Tested Claude Sonnet 4.5 for Writing and Editing · Oct 23, 2025	0.80

Anthropic called it "the best coding model in the world. It's the strongest model for building complex agents. It's the best model at using computers"^[1][4]
"Available everywhere today" as claude-sonnet-4-5 on the Claude API and in the Claude apps; Mike Krieger said it would be the default and recommended it for "basically every use case"^[1][5]
It shipped with Claude Code checkpoints and a native VS Code extension, a context editing feature and memory tool on the API, code execution and file creation in the Claude apps, and the Claude Agent SDK, the infrastructure behind Claude Code^[1][4]
"This is the most aligned frontier model we've ever released", with lower rates of sycophancy and deception, released under AI Safety Level 3 (ASL-3) protections^[1][3][4]
It arrived less than two months after Claude Opus 4.1, with Anthropic valued at $183 billion^[4][5]

Notable features

$3 per million input tokens and $15 per million output tokens, as for Sonnet 4; 200K-token context window, 64K max output and a January 2025 reliable knowledge cutoff^[1][2]
77.2% on SWE-bench Verified, averaged over 10 trials with no test-time compute, and 82.0% with parallel test-time compute^[1]
It leads OSWorld computer use at 61.4%, where Sonnet 4 had led at 42.2% four months earlier^[1]
Anthropic observed it holding focus for more than 30 hours on complex multi-step tasks; CNBC notes Opus 4 managed seven^[1][5]
Experts in finance, law, medicine and STEM found its domain knowledge and reasoning well ahead of older models, including Opus 4.1^[1]

Benchmarks^[1]

Benchmark	Claude Sonnet 4.5	Claude Opus 4.1	Claude Sonnet 4	GPT-5	Gemini 2.5 Pro
Agentic coding (SWE-bench Verified)	77.2% / 82.0% with parallel test-time compute	74.5% / 79.4% with parallel test-time compute	72.7% / 80.2% with parallel test-time compute	72.8% (GPT-5) / 74.5% (GPT-5-Codex)	67.2%
Agentic terminal coding (Terminal-Bench)	50.0%	46.5%	36.4%	43.8%	25.3%
Agentic tool use (τ2-bench), retail	86.2%	86.8%	83.8%	81.1%	—
Agentic tool use (τ2-bench), airline	70.0%	63.0%	63.0%	62.6%	—
Agentic tool use (τ2-bench), telecom	98.0%	71.5%	49.6%	96.7%	—
Computer use (OSWorld)	61.4%	44.4%	42.2%	—	—
High school math competition (AIME 2025)	100% (python) / 87.0% (no tools)	78.0%	70.5%	99.6% (python) / 94.6% (no tools)	88.0%
Graduate-level reasoning (GPQA Diamond)	83.4%	81.0%	76.1%	85.7%	86.4%
Multilingual Q&A (MMMLU)	89.1%	89.5%	86.5%	89.4%	—
Visual reasoning (MMMU validation)	77.8%	77.1%	74.4%	84.2%	82.0%
Financial analysis (Finance Agent)	55.3%	50.9%	44.5%	46.9%	29.4%

Suggest a correction

Comments

Reviews

Sentiment over timePositive

Claude Sonnet 4.5 Released

Notable features

Benchmarks[1]

Tags

Thread

Benchmarks^[1]