Wonder Full vs GPT-6 Astra: Benchmark Comparison
Wonder Full has the higher score on 11 of 24 shared benchmarks; GPT-6 Astra leads on 13.
The largest observed score gap is 45.04 pts on SRE Bench , where GPT-6 Astra leads.
Reported ±1 standard-error ranges overlap on 9 of 22 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 | Wonder Full | GPT-6 Astra | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 64.57% ±1.07 | 63.13% ±1.19 | 1.44 pts | Reported ±1 SE ranges overlap |
| Vals RSI Index | 24.43% | 27.07% | 2.64 pts | Uncertainty comparison unavailable |
| Vals Science Index | 48.54% ±2.23 | 65.09% ±2.26 | 16.55 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 21.67% ±3.54 | 5.42% ±1.17 | 16.25 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 53.37% ±3.47 | 39.42% ±3.40 | 13.94 pts | Reported ±1 SE ranges do not overlap |
| EMB | 71.92% ±2.66 | 71.70% ±2.48 | 0.22 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 59.06% ±0.21 | 53.54% ±2.08 | 5.52 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 43.00% ±3.57 | 20.72% ±2.92 | 22.28 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 49.93% ±2.18 | 48.49% ±2.13 | 1.45 pts | Reported ±1 SE ranges overlap |
| MedScribe | 89.33% ±2.00 | 87.91% ±1.94 | 1.43 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 94.00% ±2.39 | 99.00% ±1.00 | 5.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 73.33% ±0.64 | 79.26% ±0.98 | 5.93 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 35.13% ±3.21 | 53.15% ±3.36 | 18.02 pts | Reported ±1 SE ranges do not overlap |
| Robo-bench | 10.42% | 23.33% | 12.92 pts | Uncertainty comparison unavailable |
| Terminal-Bench-Science | 37.14% ±5.82 | 62.86% ±5.82 | 25.71 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 61.03% ±4.42 | 67.74% ±4.22 | 6.71 pts | Reported ±1 SE ranges overlap |
| IOI | 79.83% ±2.99 | 100.00% ±0.00 | 20.17 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 5.50% ±1.62 | 5.00 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench Multimodal | 47.29% ±2.28 | 43.33% ±2.26 | 3.96 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 58.08% ±3.31 | 59.60% ±4.40 | 1.51 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench 1-100 | 33.98% ±5.32 | 27.64% ±4.08 | 6.33 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 86.00% ±1.93 | 89.59% ±2.17 | 3.60 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 73.63% ±5.56 | 41.07% ±2.56 | 32.56 pts | Reported ±1 SE ranges do not overlap |
| SRE Bench | 11.83% ±2.00 | 56.87% ±3.07 | 45.04 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Wonder Full average | GPT-6 Astra average |
|---|---|---|
| Legal | 37.52% | 22.42% |
| Finance | 57.99% | 48.65% |
| Healthcare | 69.63% | 68.20% |
| Math | 94.00% | 99.00% |
| Science | 39.01% | 54.65% |
| Coding | 52.39% | 56.20% |
| Cyber | 42.73% | 48.97% |
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 | Wonder Full cost | GPT-6 Astra cost | Wonder Full latency | GPT-6 Astra latency |
|---|---|---|---|---|
| Vals Index | $8.56 | $18.46 | 56m36s | 27m48s |
| Vals RSI Index | N/A | N/A | 90h00m | 90h00m |
| Vals Science Index | $10.15 | $8.03 | 1h34m | 40m37s |
| Harvey's Legal Agent Benchmark | $5.41 | $26.16 | 53m21s | 24m49s |
| Legal Research Bench | $3.60 | $10.58 | 27m34s | 24m51s |
| EMB | $4.58 | $5.83 | 28m05s | 13m24s |
| Finance Agent (v2) | $2.39 | $6.82 | 14m56s | 11m33s |
| Tax Agent Bench | $2.15 | $5.80 | 18m04s | 15m01s |
| MedCode | N/A | N/A | 3m03s | 102.86s |
| MedScribe | N/A | N/A | 115.72s | 2m34s |
| ProofBench v1.1 | $0.86 | $1.71 | 11m36s | 5m08s |
| BioMysteryBench | $3.47 | $1.72 | 26m14s | 4m46s |
| MysteryMechanism | $4.16 | $1.56 | 41m34s | 8m09s |
| Robo-bench | $19.68 | $28.48 | 2m14s | 74.88s |
| Terminal-Bench-Science | $22.81 | $20.80 | 3h35m | 1h49m |
| Code Migration | $38.41 | $44.36 | 3h53m | 58m30s |
| IOI | $7.08 | $6.50 | 1h31m | 31m40s |
| ProgramBench | $28.65 | $11.57 | 6h04m | 27m02s |
| SWE-bench Multimodal | $1.86 | $1.81 | 14m22s | 7m03s |
| Terminal-Bench 4.0 | $10.53 | $9.58 | 1h12m | 35m42s |
| Vibe Code Bench 1-100 | $19.54 | $84.72 | 1h57m | 3h26m |
| Vibe Code Bench v1.1 | $2.29 | $38.51 | 10m48s | 39m04s |
| CyberBench v1.1 | $3.63 | $1.84 | 35m27s | 4m10s |
| SRE Bench | $33.81 | $10.24 | 5h34m | 17m51s |
Results available only for Wonder Full
- LegalBench
- Public Benefits Bench v1.1
Results available only for GPT-6 Astra
- SAGE
- CUA-bench
- Time Horizon Index: KSP