UnifyBench

RankingClaude Haiku 5.5

Claude Haiku 5.5

Compare models

Data updated 8 Oct 2026

31 published benchmark measures · 0 benchmark families contribute across 0 task areas. 0 capability estimates available in the full profile. See all results ↓ · Compare published benchmarks →

Model evidence summary

This profile combines published settings. It is not a runnable configuration or a leaderboard rank. Compare measured configurations →

Performance profile

Capabilities

Adjusted comparison score · 0–100
Claude Haiku 5.5 capability radarComparison scores from 0 to 100. Unknown capabilities have no point; lines stop at gaps. Hollow points are preliminary. Exact values and support labels follow the chart.AgenticHard reasoningCodingHuman prefKnowledgeMultimodalLong context50100

Filled points are supported; hollow points are preliminary. Lines stop at unknown capabilities. Exact values and sources follow below.

AgenticNo comparable evidenceUnknown

0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

Refit without one family: No supported estimate; 13/13 scenarios unsupported. Without one publisher: No supported estimate; 15/15 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage; source sensitivity still requires inspection. Validation audit.

Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

    Hard reasoningNo comparable evidenceUnknown

    0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

    Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

    Refit without one family: No supported estimate; 6/6 scenarios unsupported. Without one publisher: No supported estimate; 16/16 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

    Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.

    Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

      CodingNo comparable evidenceUnknown

      0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

      Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

      Refit without one family: No supported estimate; 6/6 scenarios unsupported. Without one publisher: No supported estimate; 19/19 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

      Audit: family and publisher holdouts improved on neutral predictions within evaluated coverage. Validation audit.

      Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

        Human prefNo comparable evidenceUnknown

        0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

        Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

        Refit without one family: No supported estimate; 1/1 scenarios unsupported. Without one publisher: No supported estimate; 1/1 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

        Audit: cross-family and cross-publisher predictions could not be evaluated from the available comparison network. Validation audit.

        Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

          KnowledgeNo comparable evidenceUnknown

          0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

          Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

          Refit without one family: No supported estimate; 4/4 scenarios unsupported. Without one publisher: No supported estimate; 5/5 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

          Audit: almost no publisher-held-out comparisons remained estimable. Cross-source validity is unresolved. Validation audit.

          Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

            MultimodalNo comparable evidenceUnknown

            0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

            Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

            Refit without one family: No supported estimate; 8/8 scenarios unsupported. Without one publisher: No supported estimate; 8/8 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

            Audit: limited publisher-held-out coverage. Cross-source validity is unresolved. Validation audit.

            Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

              Long contextNo comparable evidenceUnknown

              0 core families; 0 direct opponents across 0 labs. No admitted core comparisons

              Estimated against the same complete panel: gpt-5.6-sol claude-opus-4-8 gemini-3.1-pro . Indirect comparisons assume performance can be summarized by one strength within this capability.

              Refit without one family: No supported estimate; 5/5 scenarios unsupported. Without one publisher: No supported estimate; 5/5 unsupported. Smoothing check: No supported estimate; 3/3 unsupported. These are sensitivity checks, not confidence intervals.

              Audit: family and publisher holdouts underperformed neutral predictions. Treat this as an exploratory estimate. Validation audit.

              Matched family win shares below describe source evidence; they are not averaged to produce the adjusted score.

                Scores estimate outcomes against a shared reference panel; they are not accuracy percentages. Sparse or disconnected evidence cannot qualify an overall profile. Open a capability to inspect its evidence.

                Reported effort · Mixed settings

                Settings reported in this model's published benchmark results, including results outside the aggregate. Effort names are provider-specific. These are not API defaults or equal compute budgets.

                Mixed settings means multiple settings occur in the evidence. Best across efforts means the source selected its best reported result across settings; it does not mean Max. Unspecified settings stay unknown. This model-summary chart combines reported settings. The leaderboard keeps identified configurations separate and excludes unknown effort.

                Inspect each result and its source ↓ · Download effort evidence
                Score contributions and missing evidence

                0 contributing families across 0 capabilities. Fixed reference panels do not change when the catalog expands.

                Capability is fitted jointly across families. Capability estimates below describe different task areas; their weighted sum is not the Capability score.

                Results without reviewed compatibility or a reference match remain in the raw evidence below. Coverage counts only contributing results.

                Model information & shareable badge
                Lab
                Anthropic
                Catalog status
                active
                Availability
                Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS; account and region restrictions may apply.
                Family
                Claude Haiku
                Released
                2026-10-07
                Context
                1,000,000 tokens
                API list price
                $0.10 input / $0.50 output per million tokens
                Prompts up to 100,000 tokens. Above 100,000 tokens, $0.50 input and $2.50 output per million tokens. Cache reads $0.01/$0.05; 5-minute cache writes $0.125/$0.625 per million tokens. Adaptive thinking defaults to Medium; supported Low, Medium, High, XHigh and Max are not inferred for observations.
                License
                proprietary
                Model card
                https://platform.claude.com/docs/en/models/haiku-5-5/overview
                Default Capability family coverage
                Documented-evidence family coverage/badge/claude-haiku-5-5.svg
                Benchmark scores & sources

                Original results, evaluation harnesses, and evidence behind this model.

                BenchmarkBucketScoreHarnessEvidenceSource-recorded date
                AA-Briefcase · 1.1Supporting evidence1578
                Reported settings & source

                Artificial Analysis independently ran evaluation; max effort; provider-republished Elo.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Secondary report; no new independent unit. GDPval-AA v2.1 is anchored to DeepSeekV4.1Flash(max)=1600; AA-Briefcase is a different benchmark.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.10.2–3; p.131 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                AA-Briefcase · 1.1Supporting evidence1372
                Reported settings & source

                Artificial Analysis independently ran evaluation; medium effort; provider-republished Elo.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Secondary report; no new independent unit. GDPval-AA v2.1 is anchored to DeepSeekV4.1Flash(max)=1600; AA-Briefcase is a different benchmark.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.10.2–3; p.131 · reviewed 2026-10-07

                		Published configuration

                Effort: Medium

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                AutomationBench-AASupporting evidence35.411015132783625%Artificial Analysis AutomationBench-AA

                Effort: Not specified

                		official board

                Supporting evidence outside the reviewed capability core

                		2026-10-08
                BenchCAD · Vision2Code 1000-file subsetSupporting evidence0.67 voxel IoU
                Reported settings & source

                Adaptive thinking at max effort; without tools; mean of five runs. Random1000-file subset of17900 Vision2Code files; execution-verified CadQuery; voxel IoU.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.9.2; pp.125–127 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                BenchCAD · Vision2Code 1000-file subsetSupporting evidence0.87 voxel IoU
                Reported settings & source

                Adaptive thinking at max effort; with tools; mean of five runs. Random1000-file subset of17900 Vision2Code files; execution-verified CadQuery; voxel IoU. Container with image files, standard libraries and cropping tool.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.9.2; pp.125–127 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                Biomedical image analysisSupporting evidence43.5%
                Reported settings & source

                API, biology safeguards disabled; adaptive thinking max effort; five attempts/problem unless explicitly noted. Scores describe underlying model capability, not deployed safeguards. One attempt/problem; score scaled relative to best public competition results.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.13.5; pp.136–139 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                Bug Hunt BenchSupporting evidence20.48%Bug Hunt Bench · max

                Effort: Not specified

                		official board

                Supporting evidence outside the reviewed capability core

                		2026-10-07
                Bug Hunt BenchSupporting evidence17.43%Bug Hunt Bench · xhigh

                Effort: Not specified

                		official board

                Supporting evidence outside the reviewed capability core

                		2026-10-07
                ChartographyMultimodal46.4%
                Reported settings & source

                Adaptive thinking at max effort; without tools; mean of five runs. 100 tasks; expert answer ranges; Gemini3.5Flash grader.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.9.1; pp.123–125 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                ChartographyMultimodal86.2%
                Reported settings & source

                Adaptive thinking at max effort; with tools; mean of five runs. 100 tasks; expert answer ranges; Gemini3.5Flash grader. Container with image files, standard libraries and cropping tool.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.9.1; pp.123–125 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                FrontierCode · 1.1 ExtendedSupporting evidence58.4%
                Reported settings & source

                Cognition evaluation in Claude Code, max effort; composite functional/code-quality score; mean of five runs per task. Main:100 hardest tasks; Extended:150 tasks.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Cognition ran the evaluation; this source republishes the result. Other effort-curve points are not numerically labeled in extracted text and are excluded.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · Table 8.1.A; section 8.3; pp.111–114 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                FrontierCode · 1.1 MainSupporting evidence46.4%
                Reported settings & source

                Cognition evaluation in Claude Code, max effort; composite functional/code-quality score; mean of five runs per task. Main:100 hardest tasks; Extended:150 tasks.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Cognition ran the evaluation; this source republishes the result. Other effort-curve points are not numerically labeled in extracted text and are excluded.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · Table 8.1.A; section 8.3; pp.111–114 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                FrontierCode · 1.1 MainSupporting evidence45.8%
                Reported settings & source

                Cognition evaluation in Claude Code, xhigh effort; composite functional/code-quality score; mean of five runs per task. Main:100 hardest tasks; Extended:150 tasks.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Cognition ran the evaluation; this source republishes the result. Other effort-curve points are not numerically labeled in extracted text and are excluded.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · Table 8.1.A; section 8.3; pp.111–114 · reviewed 2026-10-07

                		Published configuration

                Effort: Xhigh

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                FrontierSWE · 2Supporting evidence43.8%
                Reported settings & source

                Proximal agent harness; max reasoning effort; 34 tasks; five trials per task; 20-hour trial budget.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Proximal ran this evaluation; retained as the provider-republished claim.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.6; p.117 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                GDPval-AA · 2.1Agentic1620
                Reported settings & source

                Artificial Analysis independently ran evaluation; max effort; provider-republished Elo.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Secondary report; no new independent unit. GDPval-AA v2.1 is anchored to DeepSeekV4.1Flash(max)=1600; AA-Briefcase is a different benchmark.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.10.2–3; p.131 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                GDPval-AA · 2.1Agentic1277
                Reported settings & source

                Artificial Analysis independently ran evaluation; medium effort; provider-republished Elo.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission. Secondary report; no new independent unit. GDPval-AA v2.1 is anchored to DeepSeekV4.1Flash(max)=1600; AA-Briefcase is a different benchmark.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.10.2–3; p.131 · reviewed 2026-10-07

                		Published configuration

                Effort: Medium

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                GMMLUKnowledge87.8%
                Reported settings & source

                Adaptive thinking max effort; safety classifiers enabled; single trial; no tools or customized system prompts;42-language mean. Blocked/unanswered0.1% excluded from average.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.12.1; pp.134–135 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench · Length-adjustedSupporting evidence59.9%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. low effort. Length-adjusted score; one run per effort below max onOct5; five-run mean at max onOct2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Low

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench · Length-adjustedSupporting evidence60.2%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. medium effort. Length-adjusted score; one run per effort below max onOct5; five-run mean at max onOct2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Medium

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench · Length-adjustedSupporting evidence60.3%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. high effort. Length-adjusted score; one run per effort below max onOct5; five-run mean at max onOct2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: High

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench · Length-adjustedSupporting evidence61.1%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. xhigh effort. Length-adjusted score; one run per effort below max onOct5; five-run mean at max onOct2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Xhigh

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench · Length-adjustedSupporting evidence61.6%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. max effort. Length-adjusted score; one run per effort below max onOct5; five-run mean at max onOct2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Max

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench Professional · Length-adjustedSupporting evidence57.9%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. low effort. Length-adjusted score; five runs per effort. Low–xhigh onOct4; max onOct1–2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Low

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench Professional · Length-adjustedSupporting evidence59.9%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. medium effort. Length-adjusted score; five runs per effort. Low–xhigh onOct4; max onOct1–2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: Medium

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07
                HealthBench Professional · Length-adjustedSupporting evidence61.3%
                Reported settings & source

                Safety classifiers enabled; same grader across models; HealthBench/Professional graded with Claude Opus4.8; PhysicianBench with Claude Opus5. high effort. Length-adjusted score; five runs per effort. Low–xhigh onOct4; max onOct1–2.

                Provider-published lab_self_report; retained as launch evidence, not an independent board admission.

                Comparison limit: Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                Claude Haiku 5.5 System Card · section 8.11.1–3; pp.131–134 · reviewed 2026-10-07

                		Published configuration

                Effort: High

                		lab self-report

                Provider publication; exact benchmark-specific configuration and independent evaluation provenance require separate review before matched-board admission.

                		Reviewed 2026-10-07