Gemini 4 Argon vs GPT-6 Astra: Benchmark Comparison
Gemini 4 Argon has the higher score on 13 of 27 shared benchmarks; GPT-6 Astra leads on 12.
The largest observed score gap is 38.00 pts on Time Horizon Index: KSP , where GPT-6 Astra leads.
Reported ±1 standard-error ranges overlap on 8 of 24 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 | Gemini 4 Argon | GPT-6 Astra | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 68.90% ±0.97 | 63.13% ±1.19 | 5.77 pts | Reported ±1 SE ranges do not overlap |
| Vals RSI Index | 30.55% | 27.07% | 3.48 pts | Uncertainty comparison unavailable |
| Vals Science Index | 55.36% ±2.29 | 65.09% ±2.26 | 9.73 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 19.58% ±3.31 | 5.42% ±1.17 | 14.17 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 54.81% ±3.46 | 39.42% ±3.40 | 15.38 pts | Reported ±1 SE ranges do not overlap |
| EMB | 75.24% ±2.46 | 71.70% ±2.48 | 3.54 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 65.40% ±0.32 | 53.54% ±2.08 | 11.86 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 44.56% ±3.59 | 20.72% ±2.92 | 23.84 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 58.80% ±2.10 | 48.49% ±2.13 | 10.32 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 87.43% ±1.96 | 87.91% ±1.94 | 0.48 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 99.00% ±1.00 | 99.00% ±1.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| BioMysteryBench | 76.30% ±0.37 | 79.26% ±0.98 | 2.96 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 45.49% ±3.35 | 53.15% ±3.36 | 7.66 pts | Reported ±1 SE ranges do not overlap |
| Robo-bench | 0.83% | 23.33% | 22.50 pts | Uncertainty comparison unavailable |
| Terminal-Bench-Science | 44.29% ±5.98 | 62.86% ±5.82 | 18.57 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 53.65% ±3.38 | 46.37% ±3.44 | 7.28 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 68.17% ±4.35 | 67.74% ±4.22 | 0.43 pts | Reported ±1 SE ranges overlap |
| IOI | 100.00% ±0.00 | 100.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 2.50% ±1.11 | 5.50% ±1.62 | 3.00 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench Multimodal | 56.04% ±2.27 | 43.33% ±2.26 | 12.71 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 57.58% ±2.31 | 59.60% ±4.40 | 2.02 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench 1-100 | 21.20% ±3.81 | 27.64% ±4.08 | 6.44 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 91.91% ±1.90 | 89.59% ±2.17 | 2.31 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 77.86% ±5.30 | 41.07% ±2.56 | 36.78 pts | Reported ±1 SE ranges do not overlap |
| SRE Bench | 44.27% ±3.08 | 56.87% ±3.07 | 12.59 pts | Reported ±1 SE ranges do not overlap |
| CUA-bench | 4.83% | 19.17% | 14.33 pts | Uncertainty comparison unavailable |
| Time Horizon Index: KSP | 52.50% ±8.86 | 90.50% ±0.00 | 38.00 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 4 Argon average | GPT-6 Astra average |
|---|---|---|
| Legal | 37.20% | 22.42% |
| Finance | 61.73% | 48.65% |
| Healthcare | 73.11% | 68.20% |
| Math | 99.00% | 99.00% |
| Science | 41.73% | 54.65% |
| Education | 53.65% | 46.37% |
| Coding | 56.77% | 56.20% |
| Cyber | 61.07% | 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 | Gemini 4 Argon cost | GPT-6 Astra cost | Gemini 4 Argon latency | GPT-6 Astra latency |
|---|---|---|---|---|
| Vals Index | $15.68 | $18.46 | 46m33s | 27m48s |
| Vals RSI Index | $179.80 | $759.51 | 90h00m | 90h00m |
| Vals Science Index | $18.40 | $8.03 | 57m00s | 40m37s |
| Harvey's Legal Agent Benchmark | $10.55 | $26.16 | 43m55s | 24m49s |
| Legal Research Bench | $6.45 | $10.58 | 31m21s | 24m51s |
| EMB | $11.80 | $5.83 | 38m26s | 13m24s |
| Finance Agent (v2) | $4.38 | $6.82 | 16m21s | 11m33s |
| Tax Agent Bench | $3.47 | $5.80 | 16m26s | 15m01s |
| MedCode | N/A | N/A | 79.53s | 102.86s |
| MedScribe | N/A | N/A | 87.44s | 2m34s |
| ProofBench v1.1 | $0.78 | $1.71 | 6m20s | 5m08s |
| BioMysteryBench | $5.71 | $1.72 | 39m57s | 4m46s |
| MysteryMechanism | $4.59 | $1.56 | 24m55s | 8m09s |
| Robo-bench | $17.33 | $28.48 | 2m32s | 74.88s |
| Terminal-Bench-Science | $44.90 | $20.80 | 1h46m | 1h49m |
| SAGE | N/A | N/A | 63.07s | 41.63s |
| Code Migration | $57.82 | $44.36 | 2h12m | 58m30s |
| IOI | $5.31 | $6.50 | 24m19s | 31m40s |
| ProgramBench | $17.66 | $11.57 | 1h11m | 27m02s |
| SWE-bench Multimodal | $11.40 | $1.81 | 33m38s | 7m03s |
| Terminal-Bench 4.0 | $17.64 | $9.58 | 53m35s | 35m42s |
| Vibe Code Bench 1-100 | $18.66 | $84.72 | 1h43m | 3h26m |
| Vibe Code Bench v1.1 | $8.11 | $38.51 | 37m28s | 39m04s |
| CyberBench v1.1 | $5.11 | $1.84 | 19m10s | 4m10s |
| SRE Bench | $30.45 | $10.24 | 1h35m | 17m51s |
| CUA-bench | $193.78 | $1821.37 | 70.51s | 32.09s |
| Time Horizon Index: KSP | $708.63 | $4206.59 | N/A | N/A |
Results available only for Gemini 4 Argon
- LegalBench
- Public Benefits Bench v1.1
Results available only for GPT-6 Astra
None.