Gemini 3.7 Flash vs GPT-6 Luna
Gemini 3.7 Flash was superseded by Gemini 3.8 Flash. The grades below are the last ones measured, so read them as history rather than as this week's board. See the current roster.
Gemini 3.7 Flash leads overall, D+ to D. By suite, real-world issues is even at C- and GPT-6 Luna takes spec planning (D+ to D-). Take GPT-6 Luna anyway when you want the side that is faster on real-world issues and ahead on spec planning.Measured
Each agent graded on the same private suites, running headless through its own tools. Advantages are oriented so a win is a win, whichever way the metric runs.
Verdict: GPT-6 Luna vs Gemini 3.7 Flash
Gemini 3.7 Flash leads overall, D+ to D. By suite: real-world issues is even at C- (60% vs 57% capability); GPT-6 Luna takes spec planning, D+ to D- (55% vs 40% capability); Gemini 3.7 Flash takes vibe coding, C- to F (61% vs 31% capability). Gemini 3.7 Flash used 9.4% of Google AI Pro's weekly limit per pass, $0.023 per task at the listed price. GPT-6 Luna used <1% of ChatGPT Plus's weekly limit per pass; at the ~$100 tier the board defaults to that is an estimated ~<0.2% of ChatGPT Pro 5x's weekly limit (est). GPT-6 Luna is faster on real-world issues: a median task takes 1m 4s against 2m 38s. Both finished every task on it. Reasons to pick GPT-6 Luna anyway: it is faster on real-world issues (1m 4s vs 2m 38s median) and it leads on spec planning (D+ to D-).
Changed since the previous graded week: Gemini 3.7 Flash: real-world issues usage 8.5% to 8.2% of the weekly limit, spec planning D to D-, spec planning usage 1.3% to 1.2% of the weekly limit; GPT-6 Luna: real-world issues D+ to C-, spec planning C- to D+.
| Dimension | Gemini 3.7 Flash | GPT-6 Luna | Advantage |
|---|---|---|---|
| Capability | 60% | 57% | +3% (Gemini 3.7 Flash leads) |
| Median task time | 2m 38s | 1m 4s | 1m 34s faster (GPT-6 Luna leads) |
| Approx. task cost | $0.025 / task | not measured | - |
Gemini 3.7 Flash: 8.2% of Google AI Pro (49.1% / 5h, 8.2% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. GPT-6 Luna: under 1% of ChatGPT Plus (2.0% / 5h) per run. Its weekly limit holds about 6 of the week's 33.6 5h windows. Each figure is a share of that agent's own plan, so there is no delta to draw.
| Dimension | Gemini 3.7 Flash | GPT-6 Luna | Advantage |
|---|---|---|---|
| Capability | 40% | 55% | +15% (GPT-6 Luna leads) |
| Median task time | 1m 38s | 1m 20s | 18s faster (GPT-6 Luna leads) |
| Approx. task cost | $0.014 / task | <$0.011 / task | $0.002 / task lower (GPT-6 Luna leads) |
Gemini 3.7 Flash: 1.2% of Google AI Pro (7.1% / 5h, 1.2% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. GPT-6 Luna: under 1% of ChatGPT Plus per run. Its weekly limit holds about 6 of the week's 33.6 5h windows. Each figure is a share of that agent's own plan, so there is no delta to draw.
| Dimension | Gemini 3.7 Flash | GPT-6 Luna | Advantage |
|---|---|---|---|
| Capability | 61% | 31% | +30% (Gemini 3.7 Flash leads) |
| Median task time | 3m 37s | 1m 24s | 2m 14s faster (GPT-6 Luna leads) |
| Approx. task cost | $0.028 / task | <$0.046 / task | $0.018 / task lower (Gemini 3.7 Flash leads) |
Gemini 3.7 Flash: 0.6% of Google AI Pro (3.6% / 5h, 0.6% / 1w) per good solve. Its weekly limit holds 6 of the week's 33.6 5h windows. GPT-6 Luna: under 1% of ChatGPT Plus per good solve. Its weekly limit holds about 6 of the week's 33.6 5h windows. Each figure is a share of that agent's own plan, so there is no delta to draw.
Grades and deltas are comparable only within a suite version. See methodology.
Frequently asked questions
Which is better for coding, Gemini 3.7 Flash or GPT-6 Luna?
Gemini 3.7 Flash clears higher: tier 4 of 5 versus tier 2 for GPT-6 Luna on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-08-30.
Which is cheaper per task, Gemini 3.7 Flash or GPT-6 Luna?
Gemini 3.7 Flash used 9.4% of Google AI Pro's weekly limit per pass, $0.023 per task at the listed price. GPT-6 Luna used <1% of ChatGPT Plus's weekly limit per pass; at the ~$100 tier the board defaults to that is an estimated ~<0.2% of ChatGPT Pro 5x's weekly limit (est).