Claude Opus 4.8 vs GPT-5.6 Terra: Benchmark Comparison

Claude Opus 4.8 has the higher score on 13 of 18 shared benchmarks; GPT-5.6 Terra leads on 5.

The largest observed score gap is 9.80 pts on MedCode , where Claude Opus 4.8 leads.

Reported ±1 standard-error ranges overlap on 11 of 18 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.

Shared benchmark results

Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.

Benchmark Claude Opus 4.8 GPT-5.6 Terra Gap Reported uncertainty
Legal Research Bench 43.75% ±3.45 41.35% ±3.42 2.40 pts Reported ±1 SE ranges overlap
LegalBench 83.57% ±0.47 85.11% ±0.45 1.54 pts Reported ±1 SE ranges do not overlap
MortgageTax 69.91% ±0.89 67.33% ±0.93 2.58 pts Reported ±1 SE ranges do not overlap
Tax Agent Bench 30.00% ±3.30 30.04% ±3.31 0.04 pts Reported ±1 SE ranges overlap
TaxEval v2 75.63% ±0.84 76.17% ±0.84 0.53 pts Reported ±1 SE ranges overlap
MedCode 53.22% ±2.17 43.41% ±2.17 9.80 pts Reported ±1 SE ranges do not overlap
MedScribe 85.75% ±1.93 82.87% ±1.95 2.89 pts Reported ±1 SE ranges overlap
Terminal-Bench-Science 4.29% ±2.44 10.00% ±3.61 5.71 pts Reported ±1 SE ranges overlap
GPQA Diamond 92.42% ±1.98 90.91% ±1.91 1.52 pts Reported ±1 SE ranges overlap
MMLU Pro 89.58% ±0.30 86.66% ±0.33 2.93 pts Reported ±1 SE ranges do not overlap
MMMU Pro 86.59% ±0.82 86.47% ±0.82 0.12 pts Reported ±1 SE ranges overlap
LiveCodeBench 87.82% ±0.95 85.93% ±1.02 1.89 pts Reported ±1 SE ranges overlap
ProgramBench 1.00% ±0.70 0.50% ±0.50 0.50 pts Reported ±1 SE ranges overlap
SkillsBench 59.23% ±4.50 58.90% ±4.47 0.33 pts Reported ±1 SE ranges overlap
SWE-bench 88.60% ±1.42 95.40% ±0.94 6.80 pts Reported ±1 SE ranges do not overlap
Terminal-Bench 4.0 23.23% ±1.34 22.73% ±2.31 0.50 pts Reported ±1 SE ranges overlap
Vibe Code Bench v1.1 82.72% ±3.08 74.59% ±4.20 8.13 pts Reported ±1 SE ranges do not overlap
Public Benefits Bench v1.1 68.13% ±1.21 62.38% ±1.26 5.75 pts Reported ±1 SE ranges do not overlap

Performance by category

Category Claude Opus 4.8 average GPT-5.6 Terra average
Legal 63.66% 63.23%
Finance 58.52% 57.84%
Healthcare 69.49% 63.14%
Science 4.29% 10.00%
Academic 89.53% 88.01%
Coding 57.10% 56.34%
Social Mobility 68.13% 62.38%

Cost and latency

Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.

Benchmark Claude Opus 4.8 cost GPT-5.6 Terra cost Claude Opus 4.8 latency GPT-5.6 Terra latency
Legal Research Bench $2.82 $7.91 11m54s 1h10m
LegalBench N/A N/A 3.11s 2.78s
MortgageTax N/A N/A 20.81s 6.32s
Tax Agent Bench $1.69 $4.83 7m59s 1h01m
TaxEval v2 N/A N/A 70.33s 37.28s
MedCode N/A N/A 105.92s 18.41s
MedScribe N/A N/A 82.36s 35.53s
Terminal-Bench-Science $23.14 $5.19 1h55m 2h04m
GPQA Diamond N/A N/A 2m14s 36.76s
MMLU Pro N/A N/A 23.44s 8.17s
MMMU Pro N/A N/A 49.04s 13.03s
LiveCodeBench N/A N/A 2m40s 44.33s
ProgramBench $31.27 $5.38 1h32m 33m36s
SkillsBench $4.77 $1.75 15m30s 6m55s
SWE-bench $1.92 $0.40 9m27s 3m00s
Terminal-Bench 4.0 $17.14 $5.60 1h11m 31m29s
Vibe Code Bench v1.1 $26.88 $7.89 1h16m 19m12s
Public Benefits Bench v1.1 $1.89 $1.20 12m48s 19m31s

Results available only for Claude Opus 4.8

  • Vals RSI Index
  • Time Horizon Index: KSP

Results available only for GPT-5.6 Terra

  • ProofBench v1.1
  • SAGE
  • IOI
  • CyberBench v1.1
  • SRE Bench
  • Financial Coaching Bench
Model details Claude Opus 4.8 Model details GPT-5.6 Terra