Gemini 3.8 Flash vs Claude Haiku 4.5
Gemini 3.8 Flash leads overall, C- to F. Claude Haiku 4.5 is about 4.7x cheaper per task. By suite, Gemini 3.8 Flash takes real-world issues (C- to D-) and spec planning (D+ to F). Take Claude Haiku 4.5 anyway when you want the side that is cheaper per task and faster on real-world issues.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: Claude Haiku 4.5 vs Gemini 3.8 Flash
Gemini 3.8 Flash leads overall, C- to F. By suite: Gemini 3.8 Flash takes real-world issues, C- to D- (60% vs 37% capability); Gemini 3.8 Flash takes spec planning, D+ to F (50% vs 21% capability); Gemini 3.8 Flash takes vibe coding, C- to F (56% vs 22% capability). Gemini 3.8 Flash used 23.7% of Google AI Pro's weekly limit per pass, $0.057 per task at the listed price. Claude Haiku 4.5 used 1% of Claude Max 5x's weekly limit per pass, $0.012 per task at the listed price. Per task, Claude Haiku 4.5 is about 4.7x cheaper ($0.012 vs $0.057). Claude Haiku 4.5 is faster on real-world issues: a median task takes 3m 8s against 5m 21s. Both finished every task on it. Reasons to pick Claude Haiku 4.5 anyway: it is cheaper per task ($0.012 vs $0.057) and it is faster on real-world issues (3m 8s vs 5m 21s median).
Changed since W39: Gemini 3.8 Flash: real-world issues F to C-, real-world issues usage 18.0% to 21.9% of the weekly limit, spec planning usage 2.0% to 1.8% of the weekly limit, vibe coding D+ to C-, vibe coding usage 1.2% to 1.1% of the weekly limit; Claude Haiku 4.5: real-world issues D+ to D-.
| Dimension | Gemini 3.8 Flash | Claude Haiku 4.5 | Advantage |
|---|---|---|---|
| Capability | 60% | 37% | +23% (Gemini 3.8 Flash leads) |
| Median task time | 5m 21s | 3m 8s | 2m 13s faster (Claude Haiku 4.5 leads) |
| Approx. task cost | $0.067 / task | $0.015 / task | $0.052 / task lower (Claude Haiku 4.5 leads) |
Gemini 3.8 Flash: 21.9% of Google AI Pro (131.3% / 5h, 21.9% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 8.0% of Claude Max 5x (8.0% / 5h, 1.0% / 1w) per run. Its weekly limit holds about 13.8 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 3.8 Flash | Claude Haiku 4.5 | Advantage |
|---|---|---|---|
| Capability | 50% | 21% | +30% (Gemini 3.8 Flash leads) |
| Median task time | 3m 0s | 2m 8s | 52s faster (Claude Haiku 4.5 leads) |
| Approx. task cost | $0.021 / task | $0.002 / task | $0.019 / task lower (Claude Haiku 4.5 leads) |
Gemini 3.8 Flash: 1.8% of Google AI Pro (11.0% / 5h, 1.8% / 1w) per run. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 1.0% of Claude Max 5x (1.0% / 5h) per run. Its weekly limit holds about 13.8 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 3.8 Flash | Claude Haiku 4.5 | Advantage |
|---|---|---|---|
| Capability | 56% | 22% | +34% (Gemini 3.8 Flash leads) |
| Median task time | 7m 51s | 4m 4s | 3m 47s faster (Claude Haiku 4.5 leads) |
| Approx. task cost | $0.048 / task | $0.009 / task | $0.040 / task lower (Claude Haiku 4.5 leads) |
Gemini 3.8 Flash: 1.1% of Google AI Pro (6.3% / 5h, 1.1% / 1w) per good solve. Its weekly limit holds 6 of the week's 33.6 5h windows. Claude Haiku 4.5: 1.3% of Claude Max 5x (1.3% / 5h) per good solve. Its weekly limit holds about 13.8 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 3.8 Flash or Claude Haiku 4.5?
Gemini 3.8 Flash clears higher: tier 4 of 5 versus tier 2 for Claude Haiku 4.5 on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-09-27.
Which is cheaper per task, Gemini 3.8 Flash or Claude Haiku 4.5?
Gemini 3.8 Flash used 23.7% of Google AI Pro's weekly limit per pass, $0.057 per task at the listed price. Claude Haiku 4.5 used 1% of Claude Max 5x's weekly limit per pass, $0.012 per task at the listed price. Per task, Claude Haiku 4.5 is about 4.7x cheaper ($0.012 vs $0.057).