SI Index Data as of Oct 9, 2026

Claude Haiku 3.5: Price, Context, Benchmarks, and Release Details

Anthropic provisional listing · first seen Oct 8, 2026 · score computed Oct 9, 2026, 01:09 UTC
40.8
SI Score (method si-v3-retained-evidence-2)
#123 of 140 ranked
100% confidence 100 percent, Full confidence — 4 of 6 expected sources in
Coverage 80% of expected source weight · 100% confidence at 80% coverage

Pillar breakdown

Coding (weight 40 percent) 34.3
Math (weight 15 percent) 32.3
Preference (weight 15 percent) 56.9
Reasoning (weight 30 percent) 38.1

Pillar weights: reasoning 30% · math 15% · coding 40% · preference 15%. Benchmark results use fixed 0–100 scales before averaging; incomplete evidence is shrunk toward 50.

Facts

Input price / 1M
$0.80Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Output price / 1M
$4.00Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Context window
200Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Max output
8.2Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released
Oct 22, 2024models.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Open weights
Nomodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
License
not yet reported
Input modalities
text, image, pdfmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗

Benchmark results

6 results
Benchmark Raw result Normalized (0–100) Pillar
aider polyglot 28%Aider polyglotPublished source fact [variant] Aider polyglot; 225 cases; 2 attemptsPublished Dec 21, 2024 Retrieved Oct 9, 2026 · Apache-2.0
Open source ↗ 		28.0 coding
gpqa diamond 38.1%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Mar 12, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 		38.1 reasoning
lmarena text 1255.3 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 2, 2026 Retrieved Oct 9, 2026 · CC-BY-4.0
Open source ↗ 		56.9 preference
math level 5 46.4%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Mar 12, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 		46.4 math
otis mock aime 2024 2025 4.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Feb 25, 2025 Retrieved Oct 9, 2026 · CC-BY
Open source ↗ 		4.3 math
swe bench verified 40.6%SWE-bench VerifiedSWE-bench published model plus agent result; harness retained, not a base model evaluation [variant] ToolsPublished Oct 22, 2024 Retrieved Oct 9, 2026 · factual citation
Open source ↗ 		40.6 coding

Normalization uses fixed absolute 0–100 scales for each unit, independently of other models that have a result on each benchmark. Coverage and evidence breadth still affect the composite; compare the evaluation conditions before reading a small score gap as decisive. “lab-reported” marks the provider's own published figure.

Sources: in and pending

in Reported (4)

  • Aider polyglot · arrived Oct 8, 2026
  • Epoch AI Benchmarking · arrived Oct 8, 2026
  • LMArena / Arena · arrived Oct 8, 2026
  • SWE-bench Verified · arrived Oct 8, 2026

pending Awaiting (2)

  • Official model cards via models.dev · carries 5% of expected weight
  • LiveBench · carries 15% of expected weight

The confidence % rises as pending sources publish. Some sources never cover some models — that is why 100% confidence arrives at 80% of expected weight, not at full coverage.