Models / Claude Sonnet 4.5

Anthropic

Claude Sonnet 4.5

Claude Sonnet 4.5 is best for writing; pick it when quality matters more than cost. The cheapest model of similar intelligence is Gemma 4 26B A4B, at $0.14 per million tokens against $6.00.

Superseded. Claude Sonnet 5.5 is Anthropic's newer model in this line, at $4.00 blended: see Claude Sonnet 5.5 or compare the two.

Retiring 30 Nov 2026. Anthropic retires claude-sonnet-4-5-20250929; the official replacement is claude-sonnet-5-5. How to re-test before you switch.

Results 16 Nov 2026 · 77 results · 10 sources

Test Claude Sonnet 4.5 on your prompt → It runs beside the models you could use instead, with every answer checked and costed.

Price per million tokens
$3.00 in · $15.00 out
Typical host on OpenRouter, 10 Nov 2026 · compare prices
Speed
Not measured
Intelligence Index
20.7
Artificial Analysis, at thinking reasoning; 19.3–20.7 across 2 settings
Spring Prompt overall
Not ranked yet
needs our benchmarks and two groups of results
Developer
Anthropic, based in the US
Weights not published
Released
29 Sep 2025
1,000,000 tokens of context

Where it stands among the models you could choose

Available models grouped by price, intelligence, speed and developer location. Retired models are excluded.

AmongIntelligence IndexSpring Prompt overallPrice, blendedSpeed
Models available today 96th of 185Claude Opus 5.5 – $6.00 blended, 90th percentilecheapest: Mistral Nemo –
Models priced $3–10 per million tokens 29th of 38Claude Opus 5.5 – sets this group –
Models with similar intelligence sets this group – $6.00 blended, 93rd percentilecheapest: Gemma 4 26B A4B –
Models from developers based in the US 53rd of 92Claude Opus 5.5 – $6.00 blended, 81st percentilecheapest: Gemma 3 4B –

The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.

How good is it, and for what?

Each use case ranks the available models its benchmarks measured, then those priced $3–10 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.

Product listingsTurning a sparse product feed and photos into listings that can go live Not measured
  • CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is Not measured
  • DeckBench: has not measured this model.
User surveysPlanning a user survey and reading its results without being misled Not measured
  • SurveyBench: has not measured this model.
Marketing planningPlanning a year of ad spend without overspending Not measured
  • ROASBench: has not measured this model.
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 103rd of 159 available · 34th of 35 at its price
BenchmarkClaude Sonnet 4.5Rank among availableBest available
tau2-benchtau2 v1.0.1, retail: task success (pass^1) · reported 72.4% 6th of 6 Qwen3.5-397B-A17B 84.4%
tau2-benchtau2 v1.0.1, airline: task success (pass^1) · reported 72.0% 6th of 6 Claude Opus 4.5 84.0%
tau2-benchtau2 v1.0.1, telecom: task success (pass^1) · reported 84.9% 6th of 6 Qwen3.5-397B-A17B 97.8%
Berkeley Function Calling Leaderboard (BFCL) V4BFCL: overall accuracy · reported 73.2% 2nd of 36 Claude Opus 4.5 77.5%
OpenHands IndexOpenHands Index: average · reported 53.0% 15th of 31 Claude Fable 5 81.0%
Artificial AnalysisArtificial Analysis: Terminal-Bench 4.0 · reported 0.0% 67th of 107 Kimi K3.1 Pro 67.2%
Artificial AnalysisArtificial Analysis: Terminal-Bench 2.1 · reported 55.8% 51st of 107 Claude Fable 5.1 91.4%
Artificial AnalysisArtificial Analysis: τ-bench banking · reported 24.5% 38th of 104 Grok 4.6 50.7%
Vending-Bench 2Vending-Bench 2: money after a year · reported $3,839 32nd of 53 GPT-6 Astra $15,515
  • Microsoft STATE-Bench: has not measured this model.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 40th of 51 available · 20th of 23 at its price
BenchmarkClaude Sonnet 4.5Rank among availableBest available
Remote Labor IndexRemote Labor Index: projects done to client standard · reported 2.1% 10th of 12 GPT-6 Astra 20.8%
  • APEX-Agents: has not measured this model.
  • GDP.pdf: has not measured this model.
Reasoning and knowledgeHard questions across science, maths and general knowledge 104th of 187 available · 31st of 39 at its price
BenchmarkClaude Sonnet 4.5Rank among availableBest available
Artificial AnalysisArtificial Analysis: Artificial Analysis Intelligence Index · reported 20.7 96th of 185 Claude Opus 5.5 57.6
Artificial AnalysisArtificial Analysis: Humanity's Last Exam · reported 17.8% 108th of 186 Claude Opus 5.5 61.4%
Artificial AnalysisArtificial Analysis: GPQA Diamond · reported 83.4% 81st of 168 GPT-6 Astra 96.3%
WritingWhat people prefer in blind comparisons, and judged writing quality 35th of 169 available · 15th of 34 at its price
BenchmarkClaude Sonnet 4.5Rank among availableBest available
Arena (formerly LMArena)Arena Text: overall · reported 1,457 43rd of 153 Gemini 4 Argon 1,525
UGI LeaderboardUGI: writing score · reported 66.9 19th of 124 Gemini 3.8 Flash 78.6
Sticking to the factsSummarising without inventing things, and factual answers 94th of 139 available · 27th of 31 at its price
BenchmarkClaude Sonnet 4.5Rank among availableBest available
Vectara Hallucination LeaderboardVectara: hallucination rate · reported 12.0% 54th of 71 GPT-5.4 nano 3.1%
Arena (formerly LMArena)Arena Text factuality: overall · reported 1,449 50th of 109 Gemini 4 Argon 1,511
SimpleQA Verified (Epoch AI)SimpleQA Verified: correct answers · reported 30.7% 54th of 68 GPT-6 Astra 75.6%
Speed under pressureGood decisions against a real clock (fast chess) Not measured
  • BulletBench: has not measured this model.

Nearest alternatives

Against Claude Sonnet 4.5 at $6.00 blended, Intelligence Index 20.7. "Similar intelligence" means within 4 points or better.

Where to run it

7 hosts on OpenRouter at the standard service tier, with a median of $6.00 blended per million tokens. Cheapest first.

HostQuantisationInputOutputBlended
Amazon Bedrocknot stated$3.00$15.00$6.00
Anthropicnot stated$3.00$15.00$6.00
Azurenot stated$3.00$15.00$6.00
Claude Platform on AWSnot stated$3.00$15.00$6.00
Googlenot stated$3.00$15.00$6.00
Amazon Bedrocknot stated$3.30$16.50$6.60
Googlenot stated$3.30$16.50$6.60

Prices per million tokens from OpenRouter's public endpoint list, checked 10 Nov 2026.

Against the alternatives

The models you would most likely weigh it against: the leaders of the groups above.

Scroll sideways to see every alternative.

Claude Sonnet 4.5Claude Opus 5.5
most intelligent at $3–10 per million tokens
Gemma 4 26B A4B
cheapest with similar intelligence
Claude Fable 5.2
leads overall
GPT-6.1 Sol
near the top overall
Price per million tokens $6.00 $8.00$0.14$16.00$4.00
Tokens a second – 9682–53
Intelligence Index 20.7 57.616.7–51.8
Spring Prompt overall – 82–9080
Rank among available models, by use case
Product listings – 13th–5th2nd
Decks from an analysis – 7th–4th5th
User surveys – 4th–1st3rd
Marketing planning – 9th–1st–
Agents and tool use 103rd of 159 3rd105th–6th
Professional work 40th of 51 3rd––9th
Reasoning and knowledge 104th of 187 1st110th–6th
Writing 35th of 169 5th82nd–16th
Sticking to the facts 94th of 139 3rd43rd–18th
Speed under pressure – ––8th3rd

Shaded figures are better than Claude Sonnet 4.5's. A dash means no result.

Research and examples

Where it has been measured

29th most intelligent of the 38 models priced $3–10 per million tokens. Among models available today, in the top quarter (by percentile) for writing; in the bottom quarter for professional work.

Artificial Analysis lists it but has not published a speed for it yet

Every result

77 published results from 10 sources, each in the source's own units, with the configuration that produced it.

Arena (formerly LMArena) · reported by Arena (formerly LMArena) · 16 results
MeasureValueRankConfigurationDated
Arena Document: overallArena rating, higher is better 1,449
1,443–1,455
33 of 44 Claude Sonnet 4.5 13 Sep 2026
Arena Search: overallArena rating, higher is better 1,175
1,170–1,180
23 of 34 Claude Sonnet 4.5 (arena search) 24 Aug 2026
Arena Text factuality: overallArena rating, higher is better 1,449
1,446–1,451
70 of 182 Claude Sonnet 4.5 2 Oct 2026
Arena Text factuality: overallArena rating, higher is better 1,447
1,444–1,449
76 of 182 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: business, management and financeArena rating, higher is better 1,461
1,455–1,466
59 of 406 Claude Sonnet 4.5 2 Oct 2026
Arena Text: business, management and financeArena rating, higher is better 1,459
1,453–1,464
65 of 406 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,454
1,448–1,460
33 of 411 Claude Sonnet 4.5 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,450
1,444–1,456
37 of 411 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,483
1,475–1,491
67 of 363 Claude Sonnet 4.5 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,498
1,490–1,506
49 of 363 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,462
1,457–1,466
46 of 413 Claude Sonnet 4.5 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,464
1,459–1,468
45 of 413 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,455
1,452–1,458
70 of 413 Claude Sonnet 4.5 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,457
1,454–1,459
65 of 413 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,454
1,449–1,459
46 of 412 Claude Sonnet 4.5 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,452
1,447–1,457
51 of 412 Claude Sonnet 4.5 (high-32k reasoning) 2 Oct 2026

Not shown: Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.

Artificial Analysis · reported by Artificial Analysis · 19 results
MeasureValueRankConfigurationDated
Artificial Analysis: Artificial Analysis Coding Indexindex score, higher is better 52.1 95 of 187 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 19.3 217 of 413 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 20.7 203 of 413 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 72.7% 233 of 360 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 83.4% 147 of 360 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 7.2% 310 of 419 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 17.8% 217 of 419 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 42.7% 189 of 288 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 57.3% 113 of 288 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 54.0% 266 of 407 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 72.3% 178 of 407 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: SciCode% of problems, higher is better 45.7% 143 of 205 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 2.1% of tasks, higher is better 55.8% 96 of 188 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 4.0% of tasks, higher is better 0.0% 152 of 204 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 28.8% 112 of 282 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 35.6% 71 of 282 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: τ-bench banking% of tasks, higher is better 24.5% 77 of 175 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 70.5% 127 of 286 Claude Sonnet 4.5 (no reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 78.1% 107 of 286 Claude Sonnet 4.5 (thinking reasoning) 20 Oct 2026

Not shown: Not your codebase or tools.

Berkeley Function Calling Leaderboard (BFCL) V4 · reported by Berkeley Function Calling Leaderboard (BFCL) V4 · 16 results
MeasureValueRankConfigurationDated
BFCL: irrelevance detection% correct, higher is better 86.6% 26 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: irrelevance detection% correct, higher is better 95.0% 5 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: memory% correct, higher is better 65.0% 2 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: memory% correct, higher is better 5.4% 89 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: multi-turn tasks% correct, higher is better 61.4% 9 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: multi-turn tasks% correct, higher is better 1.6% 97 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: overall accuracy% correct, higher is better 73.2% 2 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: overall accuracy% correct, higher is better 24.9% 89 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: relevance detection% correct, higher is better 68.8% 80 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: relevance detection% correct, higher is better 37.5% 103 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: single-turn calls (curated)% correct, higher is better 88.7% 13 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: single-turn calls (curated)% correct, higher is better 59.8% 97 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: single-turn calls (user-contributed)% correct, higher is better 81.1% 7 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: single-turn calls (user-contributed)% correct, higher is better 46.6% 102 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025
BFCL: web search% correct, higher is better 81.0% 6 of 109 Claude Sonnet 4.5 (bfcl fc) 16 Dec 2025
BFCL: web search% correct, higher is better 16.0% 47 of 109 Claude Sonnet 4.5 (bfcl prompt) 16 Dec 2025

Not shown: Not general refusal or safety behaviour.

OpenHands Index · reported by OpenHands Index · 6 results
MeasureValueRankConfigurationDated
OpenHands Index: average% resolved, higher is better 53.0% 15 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026
OpenHands Index: front end (SWE-Bench Multimodal)% resolved, higher is better 36.8% 11 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026
OpenHands Index: greenfield (Commit0)% resolved, higher is better 12.5% 25 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026
OpenHands Index: information gathering (GAIA)% resolved, higher is better 72.7% 10 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026
OpenHands Index: issue resolution (SWE-Bench)% resolved, higher is better 74.2% 17 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026
OpenHands Index: testing (SWT-Bench)% resolved, higher is better 68.8% 16 of 34 Claude Sonnet 4.5 (openhands) 30 Jun 2026

Not shown: Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.

Remote Labor Index · reported by Remote Labor Index · 1 result
MeasureValueRankConfigurationDated
Remote Labor Index: projects done to client standard% of projects, higher is better 2.1% 10 of 14 Claude Sonnet 4.5 1 Oct 2026

Not shown: Not speed or cost; a small set of projects, so a few points either way are noise.

SimpleQA Verified (Epoch AI) · reported by SimpleQA Verified (Epoch AI) · 2 results
MeasureValueRankConfigurationDated
SimpleQA Verified: correct answers% of questions, higher is better 23.7% 69 of 85 Claude Sonnet 4.5 10 Aug 2026
SimpleQA Verified: correct answers% of questions, higher is better 30.7% 64 of 85 Claude Sonnet 4.5 (thinking-59k reasoning) 27 Aug 2026

Not shown: Not answers grounded in your documents; tests what the model remembers.

tau2-bench · reported by tau2-bench · 8 results
MeasureValueRankConfigurationDated
tau2 v1.0.1, airline: consistency (pass^4)% of tasks, higher is better 48.0% 7 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, airline: task success (pass^1)% of tasks, higher is better 72.0% 7 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, banking knowledge, terminal retrieval: consistency (pass^4)% of tasks, higher is better 10.3% 1 of 3 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, banking knowledge, terminal retrieval: task success (pass^1)% of tasks, higher is better 25.3% 2 of 3 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, retail: consistency (pass^4)% of tasks, higher is better 39.5% 8 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, retail: task success (pass^1)% of tasks, higher is better 72.4% 8 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, telecom: consistency (pass^4)% of tasks, higher is better 64.0% 6 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026
tau2 v1.0.1, telecom: task success (pass^1)% of tasks, higher is better 84.9% 7 of 8 Claude Sonnet 4.5 (thinking reasoning, tau2) 24 Feb 2026

Not shown: Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.

UGI Leaderboard · reported by UGI Leaderboard · 6 results
MeasureValueRankConfigurationDated
UGI: requested-length error% off the requested word count, lower is better 5.0% 41 of 379 Claude Sonnet 4.5 (no reasoning) 30 Sep 2025
UGI: requested-length error% off the requested word count, lower is better 8.0% 63 of 379 Claude Sonnet 4.5 (thinking reasoning) 30 Sep 2025
UGI: style adherencescore from 0 to 1, higher is better 0.38 75 of 379 Claude Sonnet 4.5 (no reasoning) 30 Sep 2025
UGI: style adherencescore from 0 to 1, higher is better 0.41 16 of 379 Claude Sonnet 4.5 (thinking reasoning) 30 Sep 2025
UGI: writing scorescore out of 100, higher is better 66.9 50 of 379 Claude Sonnet 4.5 (no reasoning) 30 Sep 2025
UGI: writing scorescore out of 100, higher is better 64.2 75 of 379 Claude Sonnet 4.5 (thinking reasoning) 30 Sep 2025

Not shown: Not other format limits such as character counts or bullet counts.

Vectara Hallucination Leaderboard · reported by Vectara Hallucination Leaderboard · 2 results
MeasureValueRankConfigurationDated
Vectara: answer rate% of documents, higher is better 95.6% 90 of 108 Claude Sonnet 4.5 22 Sep 2026
Vectara: hallucination rate% of summaries, lower is better 12.0% 78 of 108 Claude Sonnet 4.5 22 Sep 2026

Not shown: Not a quality score: a low rate usually means content filters were triggered, and hallucination rates are measured on answered documents only.

Vending-Bench 2 · reported by Vending-Bench 2 · 1 result
MeasureValueRankConfigurationDated
Vending-Bench 2: money after a yearUS dollars, higher is better $3,839 38 of 63 Claude Sonnet 4.5 1 Oct 2026

Not shown: Not a real business; one simulated market with set rules.

Will Claude Sonnet 4.5 work for your product?

Benchmarks aren't your task. Paste the prompt your product uses and run it beside the models you could use instead, with every answer checked and costed.

New here? Your email starts a free trial. No credit card needed.