Claude Haiku 4.5 vs GPT-5.6 Luna
GPT-5.6 Luna was superseded by GPT-6 Luna. The grades below are the last ones measured, so read them as history rather than as this week's board. See the current roster.
GPT-5.6 Luna leads overall, D to F. By suite, GPT-5.6 Luna takes real-world issues (D to D-) and spec planning (C- to F). The data gives no reason to pick Claude Haiku 4.5 over GPT-5.6 Luna this week. Both agents run the same private suite cut headless through their own tools, each on its own subscription.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-5.6 Luna vs Claude Haiku 4.5
GPT-5.6 Luna leads overall, D to F. By suite: GPT-5.6 Luna takes real-world issues, D to D- (45% vs 37% capability); GPT-5.6 Luna takes spec planning, C- to F (57% vs 21% capability); vibe coding is even at F (22% vs 25% capability). Claude Haiku 4.5 used 1% of Claude Max 5x's weekly limit per pass, $0.012 per task at the listed price. GPT-5.6 Luna used <1% of ChatGPT Pro 5x's weekly limit per pass, at most $0.012 per task at the listed price. GPT-5.6 Luna is faster on real-world issues: a median task takes 3m 7s against 3m 8s. Both finished every task on it. The data gives no reason to pick Claude Haiku 4.5 over GPT-5.6 Luna this week.
Changed since the previous graded week: Claude Haiku 4.5: real-world issues D+ to D-; GPT-5.6 Luna: spec planning D+ to C-, vibe coding C- to F.
| Dimension | Claude Haiku 4.5 | GPT-5.6 Luna | Advantage |
|---|---|---|---|
| Capability | 37% | 45% | +8% (GPT-5.6 Luna leads) |
| Median task time | 3m 8s | 3m 7s | 1s faster (GPT-5.6 Luna leads) |
| Approx. task cost | $0.015 / task | <$0.015 / task | Even |
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. GPT-5.6 Luna: under 1% of ChatGPT Pro 5x per run. Each figure is a share of that agent's own plan, so there is no delta to draw.
| Dimension | Claude Haiku 4.5 | GPT-5.6 Luna | Advantage |
|---|---|---|---|
| Capability | 21% | 57% | +36% (GPT-5.6 Luna leads) |
| Median task time | 2m 8s | 3m 25s | 1m 17s faster (Claude Haiku 4.5 leads) |
| Approx. task cost | $0.002 / task | <$0.057 / task | $0.056 / task lower (Claude Haiku 4.5 leads) |
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. GPT-5.6 Luna: under 1% of ChatGPT Pro 5x per run. Each figure is a share of that agent's own plan, so there is no delta to draw.
| Dimension | Claude Haiku 4.5 | GPT-5.6 Luna | Advantage |
|---|---|---|---|
| Capability | 22% | 25% | +3% (GPT-5.6 Luna leads) |
| Median task time | 4m 4s | 3m 40s | 24s faster (GPT-5.6 Luna leads) |
| Approx. task cost | $0.009 / task | <$0.230 / task | $0.221 / task lower (Claude Haiku 4.5 leads) |
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. GPT-5.6 Luna: under 1% of ChatGPT Pro 5x per good solve. 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, Claude Haiku 4.5 or GPT-5.6 Luna?
Claude Haiku 4.5 clears higher: tier 2 of 5 versus tier 1 for GPT-5.6 Luna on real-world@v3, our private suite of real-world coding problems. Latest measurement 2026-09-28.
Which is cheaper per task, Claude Haiku 4.5 or GPT-5.6 Luna?
Claude Haiku 4.5 used 1% of Claude Max 5x's weekly limit per pass, $0.012 per task at the listed price. GPT-5.6 Luna used <1% of ChatGPT Pro 5x's weekly limit per pass, at most $0.012 per task at the listed price.
What do Claude Haiku 4.5 and GPT-5.6 Luna cost a month?
Claude Haiku 4.5 runs on Claude Max 5x at $100/month (public price as of 2026-07-20). GPT-5.6 Luna runs on ChatGPT Pro 5x at $100/month (public price as of 2026-09-05).