Home › Videos › Claude Haiku 5.5 vs GPT-6 Luna — Same Price, 4.5× the Bill?

8 Oct 2026 · 2 runs · 2 models

Claude Haiku 5.5 vs GPT-6 Luna — Same Price, 4.5× the Bill?

Claude Haiku 5.5 and GPT-6 Luna list at the same price per token. On the same eight builds they scored 6.1 and 6.2, but Haiku wrote 4.5× the output tokens, so its bill was at least $3.79 against Luna's $0.85.

Watch on YouTube Compare models from this video Every score (CSV)

The results

The same scores as the leaderboard: a test that failed is left out of its run's score, not counted as zero. Time, tokens and cost cover each whole run, wildcards included.

See the whole board

Ranking

Quality v1 over measured non-wildcard pages. Different workloads: browse the outputs and coverage before comparing models.

gpt-6-lunaeffort: maxOpenAI · 8/8 scored pages

6.2

claude-haiku-5-5effort: highAnthropic · 8/8 scored pages

6.1

0510

Score against output tokens

Quality v1 against full-run tokens. Different workloads and hardware affect this comparison.

Claude HaikuGPT Luna

Output tokens include the model's thinking. Like the time, tokens and cost table, it counts every test the model took, wildcards included. A hollow dot is not an exact count: some tests reported none, or the count was worked out from the length of the output. The score axis is zoomed to the data.

Score against time

Quality v1 against full-run time. Different workloads and hardware affect this comparison.

Claude HaikuGPT Luna

Time is how long the model took to answer, added up over its tests; judging is not counted. Like the time, tokens and cost table, it counts every test the model took, wildcards included. A hollow dot is a lower bound: some tests have no time recorded. The score axis is zoomed to the data.

Where each model is strong and weak

One row per model, one column per test and per scoring criterion.

	By test									By what the judges scored
Model	Aero Lab	Jet Odyssey	Blast Island	Launch Day	Alex Optics	Afterburner	Hangar Ops	Hangar Planner	Rotor Rampage	Working behaviour	Visual quality	Engineering quality	Usability / accessibility
gpt-6-lunaeffort: max	7.3	7.3	1.4	5.8	7.7	7.0	7.5	—	5.7	6.5	6.7	6.0	5.2
claude-haiku-5-5effort: high	6.9best of 4	1.4best of 3	6.8best of 3	7.1best of 2	7.8	7.0best of 2	—	7.2	4.7best of 2	6.3	6.6	6.2	5.0

Violet is strong, orange is weak, on the same 0–10 scale in every cell.Different prompts and protocols: no best-in-column verdict.

Time, tokens and cost

What each model used across all its tests, wildcards included.

Model	Tests	Time	Output tokens	Speed	Cost	Subscription used
gpt-6-lunaeffort: maxOpenAI	8	3h 38m	577k321k of it thinking	43.9 tok/s	≈$0.85API price, est.	7-day window +5 pts
claude-haiku-5-5effort: highAnthropic	9	≥4h 14m	≥2.58M2.15M of it thinking	222.3 tok/s	≥$3.79mixed	—

≈ An estimate: the provider reported no count, so it is worked out from the length of the output (about 4 characters a token). — Not measured. Cost says how it was worked out: charged by the provider, at API prices (for a subscription), or no API charge on local hardware. ≥ means some tests had no price.

Read from the account's usage window in whole percentage points, so a short run can show 0 points. It counts everything the account did in that time, not only this run.

Runs in this video

Every run behind the numbers. Scores name their comparison basis.

Model

Score / basis

Time

Output tokens

Cost

gpt-6-lunaeffort: maxOpenAI · 2 Oct 2026

6.2 / 10Quality v1 · 8/8 pages

Time3h 38m

Output tokens577k

Cost≈$0.85API price, est.

Open Compare

claude-haiku-5-5effort: highAnthropic · 7 Oct 2026

6.1 / 10Quality v1 · 8/8 pages

Time≥4h 14m

Output tokens≥2.58M

Cost≥$3.79mixed

Open Compare

gpt-6-lunaeffort: max

OpenAI · 6.2 / 10 · Quality v1 · 8/8 pages · 3h 38m · ≈$0.85

metrics.csv run.json Open the run page Compare

Screenshot of the page gpt-6-luna made for Aero Lab

7.3

Aero Lab gpt-6-luna · effort: max 38m 41s · 94k tokens · ≈$0.12 Quality v1 · Working behaviour 7.0 · Visual quality 8.0 · Engineering quality 7.0 · Usability / accessibility 7.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 3ef604ea3997fb42

Earlier evaluations

5.3/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Jet Odyssey

7.3

Jet Odyssey gpt-6-luna · effort: max 20m 43s · 66k tokens · ≈$0.11 Quality v1 · Working behaviour 8.0 · Visual quality 8.0 · Engineering quality 6.5 · Usability / accessibility 6.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 14bd7252a809672d

Earlier evaluations

5.2/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Blast Island

1.4

Blast Island gpt-6-luna · effort: max 28m 59s · 93k tokens · ≈$0.17 Quality v1 · Working behaviour 0.5 · Visual quality 1.5 · Engineering quality 3.0 · Usability / accessibility 1.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt a282381696456fc9

Earlier evaluations

1.5/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Launch Day

5.8

Launch Day gpt-6-luna · effort: max 1h 01m · 146k tokens · ≈$0.26 Quality v1 · Working behaviour 6.5 · Visual quality 6.0 · Engineering quality 5.5 · Usability / accessibility 4.5scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 155b76cde8ba1880

Earlier evaluations

3.8/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Alex Optics

7.7

Alex Optics gpt-6-luna · effort: max 15m 39s · 42k tokens · ≈$0.03 Quality v1 · Working behaviour 8.0 · Visual quality 8.5 · Engineering quality 7.0 · Usability / accessibility 6.5scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt e519ec9dadc0a186

Earlier evaluations

6.5/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Afterburner

7.0

Afterburner gpt-6-luna · effort: max 21m 06s · 69k tokens · ≈$0.09 Quality v1 · Working behaviour 7.5 · Visual quality 7.5 · Engineering quality 6.5 · Usability / accessibility 6.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt e4ab4889f06802b1

Earlier evaluations

6.2/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Hangar Ops

7.5

Hangar Ops gpt-6-luna · effort: max 5m 23s · 17k tokens · ≈$0.01 Quality v1 · Working behaviour 8.0 · Visual quality 7.5 · Engineering quality 7.5 · Usability / accessibility 6.5Fixed 9 of 10 bugs · 8 of 8 features still work · 105 lines changedscored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 86adb1d1b1c125c3

Earlier evaluations

8.3/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

Screenshot of the page gpt-6-luna made for Rotor Rampage

5.7

Rotor Rampage gpt-6-luna · effort: max 27m 22s · 49k tokens · ≈$0.07 Quality v1 · Working behaviour 6.5 · Visual quality 6.5 · Engineering quality 5.0 · Usability / accessibility 4.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 2db3f721e5dfca4a

Earlier evaluations

5.9/10 · engineering-1 · claude-code/claude-opus-5-5 (engineering-1 rubric)

Different rubrics; these values are historical, not a score change.

Open the page Details

claude-haiku-5-5effort: high

Anthropic · 6.1 / 10 · Quality v1 · 8/8 pages · 4h 14m · ≥$3.79

metrics.csv run.json Open the run page Compare

Screenshot of the page claude-haiku-5-5 made for Aero Lab

6.9

Aero Labbest of 4 claude-haiku-5-5 · effort: high 11m 08s · 157k tokens · ≈$0.50 Quality v1 · Working behaviour 7.5 · Visual quality 6.5 · Engineering quality 7.5 · Usability / accessibility 6.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 1e8e809e8caeb2ba Open the page Details

Screenshot of the page claude-haiku-5-5 made for Jet Odyssey

1.4

Jet Odysseybest of 3 claude-haiku-5-5 · effort: high 44m 53s · 594k tokens · $0.83 Quality v1 · Working behaviour 0.5 · Visual quality 1.5 · Engineering quality 2.5 · Usability / accessibility 1.5scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 1ce617b6d67089b2 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Blast Island

6.8

Blast Islandbest of 3 claude-haiku-5-5 · effort: high 21m 28s · 284k tokens · $0.34 Quality v1 · Working behaviour 7.5 · Visual quality 7.5 · Engineering quality 6.5 · Usability / accessibility 5.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 28ba22cd0f56ca1e Open the page Details

Screenshot of the page claude-haiku-5-5 made for Launch Day

7.1

Launch Daybest of 2 claude-haiku-5-5 · effort: high 49m 35s · 636k tokens · $0.95 Quality v1 · Working behaviour 8.0 · Visual quality 7.5 · Engineering quality 6.0 · Usability / accessibility 6.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt d1e76d48a7ac8eb6 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Alex Optics

7.8

Alex Optics claude-haiku-5-5 · effort: high 7m 29s · 105k tokens · $0.17 Quality v1 · Working behaviour 8.2 · Visual quality 8.4 · Engineering quality 7.6 · Usability / accessibility 6.6scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt de7c304e41dd3505 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Afterburner

7.0

Afterburnerbest of 2 claude-haiku-5-5 · effort: high 19m 50s · 270k tokens · $0.36 Quality v1 · Working behaviour 7.5 · Visual quality 7.5 · Engineering quality 7.0 · Usability / accessibility 5.5scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt e08d0617005c33f9 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Hangar Planner

7.2

Hangar Planner claude-haiku-5-5 · effort: high 2m 27s · 41k tokens · $0.03 Quality v1 · Working behaviour 7.5 · Visual quality 7.5 · Engineering quality 8.0 · Usability / accessibility 5.5Fixed 16 of 16 bugs · 14 of 14 features still work · 112 lines changedscored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 14deca81c8b3a6c9 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Rotor Rampage

4.7

Rotor Rampagebest of 2 claude-haiku-5-5 · effort: high 9m 44s · 137k tokens · $0.25 Quality v1 · Working behaviour 3.5 · Visual quality 6.5 · Engineering quality 4.5 · Usability / accessibility 4.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt d196c18e5127f818 Open the page Details

Screenshot of the page claude-haiku-5-5 made for Cold Shock Parison

7.2

Cold Shock Parisonwildcard claude-haiku-5-5 · effort: high 7m 01s · 98k tokens · $0.07 Quality v1 · Working behaviour 8.0 · Visual quality 7.5 · Engineering quality 7.5 · Usability / accessibility 5.0scored by claude-code/claude-opus-5-5 (quality-1 rubric)prompt 1de77f2693bd93ba Open the page Details