sugra.blog

BlogLLM Benchmark Scores as Published, Side by Side

Claude Haiku 4.5: Benchmark Scores as Published

Every result the catalog holds for Claude Haiku 4.5: 113 tests in 14 areas, from 6 pages. 106 printed by Anthropic itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.

Developer AnthropicReleased 2025-10-15Context window 200KMax output 64KInput Text, imageOutput TextWeights ClosedData version 2026103001, as of 2026-10-30

Compare in the table

Knowledge Work

TestValuePrinted bySetup as printed
GDPval-AA v2907Google, Gemini 3.5 Flash-Lite model cardchecked 2026-10-15-
OfficeQA Pro47.1%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
GDPval-AA v2.1735Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more page-
AA-Briefcase v1.1614Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more page-
OfficeQA63%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

Agentic Coding

TestValuePrinted bySetup as printed
SWE-bench Verified73.3%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
SWE-Bench Pro39.5%Google, Gemini 3.5 Flash-Lite model cardchecked 2026-10-15-
Terminal-Bench 2.144.2%Google, Gemini 3.5 Flash-Lite model cardchecked 2026-10-15Harness: Terminus-2
SWE-bench Multilingual67.4%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Terminal-Bench 4.00%Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more page-
Terminal-Bench (version not stated)41%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
SWE-Bench Multimodal19.8%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

Science and Math

TestValuePrinted bySetup as printed
GPQA Diamond73%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
AIME 202596.3%Anthropic, Claude Haiku 4.5checked 2026-10-15; printed on 1 more pageEffort: thinking budget 128K; Tools: Python
80.7%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K; Tools: none
Terminal-Bench Science 0.10.1%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

Long Context

TestValuePrinted bySetup as printed
MRCR v2 8-needle 128K35.3%Google, Gemini 3.5 Flash-Lite model cardchecked 2026-10-15Context: 128k (average)
LCR (Mistral custom implementation of AA LCR)64Mistral AI, Mistral Small 4 model cardchecked 2026-10-15Effort: instruct; Harness: Score vs. Output Length chart (Mistral comparison run)
80Mistral AI, Mistral Small 4 model cardchecked 2026-10-15Effort: reasoning; Harness: Score vs. Output Length chart (Mistral comparison run)

Computer Use

TestValuePrinted bySetup as printed
OSWorld-Verified50.7%Anthropic, Claude Haiku 4.5checked 2026-10-15; printed on 1 more pageEffort: thinking budget 128K
OSWorld 2.1 offline subset15.7%Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more page-

Multimodal Understanding

TestValuePrinted bySetup as printed
CharXiv Reasoning (no tools)61.7%Google, Gemini 3.5 Flash-Lite model cardchecked 2026-10-15Tools: none
MMMU (validation)73.2%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
Chartography (no tools)6.4%Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more pageTools: none
Chartography (with tools)8.8%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Tools: enabled

Cybersecurity

TestValuePrinted bySetup as printed
Binary Exploitation (control-flow hijacks)0Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
CyScenarioBench0%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Dangerous actions while oblivious to a possibly real settinglower is better1.21Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Motivated reasoning (cyber)lower is better5.64Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Unsanctioned sandbox escape attemptlower is better2.16Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

General Knowledge and Reasoning

TestValuePrinted bySetup as printed
Humanity's Last Exam (no tools)10.2%Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more pageTools: none
Humanity's Last Exam (with tools)18.7%Anthropic, Claude Haiku 5.5checked 2026-10-15; printed on 1 more pageTools: enabled; exact tools in source

Code Generation

TestValuePrinted bySetup as printed
LiveCodeBench (version not stated)40%Mistral AI, Mistral Small 4 model cardchecked 2026-10-15Effort: instruct; Harness: Score vs. Output Length chart (Mistral comparison run)
49%Mistral AI, Mistral Small 4 model cardchecked 2026-10-15Effort: reasoning; Harness: Score vs. Output Length chart (Mistral comparison run)

Tool Use and Workflows

TestValuePrinted bySetup as printed
τ²-bench Telecom83%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
τ²-bench Retail83.2%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K
τ²-bench Airline63.6%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K

Multilingual

TestValuePrinted bySetup as printed
MMMLU83%Anthropic, Claude Haiku 4.5checked 2026-10-15Effort: thinking budget 128K

Healthcare

TestValuePrinted bySetup as printed
HealthBench Professional (length-adjusted)32.2%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

Safety and Alignment

TestValuePrinted bySetup as printed
Gray Swan IPI (attack success rate after k attempts)attack success rate after k attempts, lower is better83.2%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
BBQ ambiguous accuracy99.11%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
BBQ disambiguated accuracy60.56%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Benign request refusal rate (all languages)lower is better3.05%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.1.2.A, all tested languages, claude.ai
0.44%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.1.2.A, all tested languages, API
Child safety multi-turn appropriate response rate95%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.B, API, without a system prompt
98%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.B, claude.ai
Child safety single-turn harmless rate99.92%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.A, API
99.92%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.A, claude.ai
Child safety single-turn refusal ratelower is better2.71%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.A, claude.ai
0.33%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.2.A, API
Claude Code dual-use and benign success rate88.7%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Claude Code malicious request refusal rate66.6%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Disordered eating single-turn harmless rate99.04%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.2.A, claude.ai
96.51%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.2.A, API
Disordered eating single-turn refusal ratelower is better0.68%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.2.A, claude.ai
0.07%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.2.A, API
Election integrity multi-turn appropriate response rate87%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.B, claude.ai
93%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.B, API
Election integrity single-turn harmless rate99.83%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.A, API
99.83%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.A, claude.ai
Election integrity single-turn refusal ratelower is better0%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.A, claude.ai
0%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Election integrity prompts (section 4.4.3), Table 4.4.3.A, API
Harmful request harmless response rate (all languages)99.32%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.1.1.A, all tested languages, claude.ai
97.23%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.1.1.A, all tested languages, API
Malicious computer use refusal rate (without mitigations)58.93%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Pairwise even-handedness86.7%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, API
91.2%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, claude.ai
Pairwise opposing perspectives acknowledged49.9%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, API
45.1%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, claude.ai
Shade coding attack success rate (attempt level)lower is better30.43%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Shade indirect prompt injection attack success (scenario counts as printed in the card), coding, with probes: H5.5 0/40, Sonnet 5.5 13/40, Opus 5.5 36/40, Fable 5.1 35/40, Haiku 4.5 40/40
58.4%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Shade indirect prompt injection attack success (scenario counts as printed in the card), coding, no safeguards: H5.5 6/40, Sonnet 5.5 16/40, Opus 5.5 37/40, Fable 5.1 39/40, Haiku 4.5 40/40
Shade computer use attack success rate (attempt level)lower is better18.89%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Shade indirect prompt injection attack success (scenario counts as printed in the card), computer use GUI, no safeguards: 1/14 for Haiku 5.5, Sonnet 5.5, Opus 5.5 and Fable 5.1; 13/14 for Haiku 4.5
10.46%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Shade indirect prompt injection attack success (scenario counts as printed in the card), computer use GUI, with probes: 1/14 for Haiku 5.5, Sonnet 5.5 and Fable 5.1; 1/14 for Opus 5.5; 10/14 for Haiku 4.5
Suicide and self-harm multi-turn appropriate response rate71%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.B, claude.ai
46%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.B, API
Suicide and self-harm single-turn harmless rate99.79%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.A, claude.ai
99.46%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.A, API
Suicide and self-harm single-turn refusal ratelower is better0.55%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.A, claude.ai
0.02%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Table 4.3.1.A, API
Behavioral audit: Accepting unverifiable authorizationlower is better4.55Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Admirable behavior3.77Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Behavior consistency4.18Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Biological weapons upliftlower is better1.046Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Causal decision theory adherencelower is better4.6Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Character driftlower is better2.078Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Chemical weapons upliftlower is better1.09Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Circumventing constraints or controlslower is better3.32Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Coherence between actions and views3.3Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Complying with prompt injectionslower is better4.08Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Condescension toward the userlower is better1.67Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Controlled substance synthesis upliftlower is better1.137Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Creative mastery5.39Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Disallowed cyberoffenselower is better1.447Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Encouragement of user delusionlower is better1.165Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Evasiveness on controversial topicslower is better1.71Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Explosive weapons upliftlower is better1.099Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Failure to disclose bad or lazy behaviorlower is better4.38Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: False completion claimslower is better4.85Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Fraudlower is better1.88Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Fun or funny behavior2.45Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Good for the user4.01Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Harmful sexual contentlower is better1.09Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Ignoring explicit constraintslower is better4.46Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Illegible thinkinglower is better1.41Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Important omissionslower is better4.74Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Input hallucinationlower is better3.96Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Intellectual depth5.25Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Military-grade weaponslower is better1.083Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Overrefusallower is better1.727Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Proposes a security shortcut in the name of expediencelower is better5.15Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Radiological or nuclear weapons upliftlower is better1.009Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Reckless tool uselower is better4.14Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Supporting user autonomy5.35Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Sycophancylower is better4.15Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Undermining liberal democracylower is better1.43Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Unfaithful thinkinglower is better4.41Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: User deceptionlower is better3.53Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Verbalized evaluation awarenesslower is better1.67Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Warmth5.8Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Behavioral audit: Wet blanketlower is better1.698Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Multi-turn appropriate response rate: Bio weapons85%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
97%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Cyberattacks94%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
97%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Deadly weapons84%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
80%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Hate and discrimination100%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
100%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Influence operations68%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
78%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Romance scams100%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
93%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Tracking and surveillance84%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
89%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Multi-turn appropriate response rate: Violent extremism93%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, API (solid bars), Figure 4.1.3.A
89%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Multi-turn, claude.ai (hatched bars), Figure 4.1.3.A
Political refusals (Pairwise political refusal rate)Pairwise political refusal rate, lower is better5.4%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, API
3.6%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Setup as reported: Pairwise political bias evaluation (Figure 4.4.1.A-C); the card prints the values as bar labels, claude.ai
Sandbox escape attempted crossings, all severitieslower is better19.9%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Sandbox escape attempted crossings, low severitylower is better4.1%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-
Sandbox escape attempted crossings, moderate severity and abovelower is better15.8%Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19-

Engineering and Design

TestValuePrinted bySetup as printed
BenchCAD Vision2Code voxel IoU0.155Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Tools: none
0.21Anthropic, Claude Haiku 5.5 System Cardchecked 2026-10-19Tools: enabled

Price

API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked.

PriceValuePrinted bySetup as printed
Input$1Anthropic, Anthropic API pricingchecked 2026-10-15-
Cached input$0.1Anthropic, Anthropic API pricingchecked 2026-10-15-
Cache write$1.25Anthropic, Anthropic API pricingchecked 2026-10-15Cache: 5 minute cache
Cache write, 1 hour$2Anthropic, Anthropic API pricingchecked 2026-10-15-
Output$5Anthropic, Anthropic API pricingchecked 2026-10-15-
Input, batch$0.5Anthropic, Anthropic API pricingchecked 2026-10-15-
Output, batch$2.5Anthropic, Anthropic API pricingchecked 2026-10-15-

More from Anthropic

Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.