# Haiku 5.5 max vs Opus 4.8 medium

Stet publishes a public search API over its indexed pages; the guide is at https://www.stet.sh/agents/agent-access.

Source: https://www.stet.sh/leaderboard/compare/haiku-5-5-max-vs-opus-4-8-medium
Source observed: 2026-11-19T11:08:36.950Z
Last checked: 2026-11-19T11:08:36.95+00:00
Indexed: 2026-11-19T11:15:21.652+00:00

---

# Haiku 5.5 max vs Opus 4.8 medium

Opus 4.8 medium leads on speed, Haiku 5.5 max leads on cost, and quality is not clear from this comparison. Both setups ran `claude-code`, with max reasoning for Haiku and medium reasoning for Opus. We compare setups, not models alone.

## Headline results

| Dimension | Result |
|---|---|
| Quality | Even: Haiku won 6, Opus won 6, 10 were tied, and 2 were undecided. |
| Speed | Opus 4.8 medium was **2.5× faster**: 23.3 vs 58.4 minutes per task. |
| Cost | Haiku 5.5 max was **2.0× cheaper**: $3.83 vs $7.68 per task. |

## What we compared

- **1 real repository:** private `Stet-AI/stet`; not public benchmark scores.
- **24 merged changes:** each setup ran end to end from the repository state before the change.
- **Grading:** the repository’s own tests and judges comparing each patch with the merged change.
- **Measures:** quality, cost, and time on real agent work; the sample is smaller than public benchmarks.

For all effort pairings, see [Haiku 5.5 vs Opus 4.8](https://www.stet.sh/leaderboard/compare/haiku-5-5-vs-opus-4-8). Data is from **November 19, 2026**.

## Stet score and quality

| Setup | Harness / reasoning | Rank | Stet score | 95% range |
|---|---|---:|---:|---:|
| Haiku 5.5 max | `claude-code` / max | 10 of 18 | 48.7 | 38.3–58.7 |
| Opus 4.8 medium | `claude-code` / medium | 13 of 18 | 46.8 | 38.0–55.7 |

The 95% score ranges overlap. Opus 4.8 medium’s win share is 52%, with a 95% range of 37–66%; ties are within judge noise.

### Other quality measures

| Metric | Haiku 5.5 max | Opus 4.8 medium |
|---|---:|---:|
| Code review pass rate | 38% (24 tasks) | 33% (24 tasks) |
| Checklist score | 0.81 (24 tasks) | 0.78 (24 tasks) |
| Equivalence rate | 42% (24 tasks) | 46% (24 tasks) |

These measures are not part of the headline Stet score and have no confidence interval; a small gap may be noise. The higher recorded value is not a proven lead.

## Task by task

Haiku solved 2 tasks that Opus did not. Opus solved 5 that Haiku did not. Both solved 8, and neither solved 7. On the 8 tasks both solved, judges preferred Haiku’s patch on 4, Opus’s on 1, and called 3 even.

### Only Haiku 5.5 max solved

- `stet-2450ca2d` — 2 changed lines in 1 file; judges preferred Haiku’s patch.
- `stet-commit-342d0ac0` — 640 changed lines in 8 files; judges preferred Haiku’s patch.

### Only Opus 4.8 medium solved

- `stet-43493c52` — 4 lines, 1 file; judges preferred Opus’s patch.
- `stet-232b50ba` — 10 lines, 1 file; judges preferred Opus’s patch.
- `stet-66762b4c` — 84 lines, 1 file; judges preferred Opus’s patch.
- `stet-commit-f0caded9` — 113 lines, 4 files; judges preferred Opus’s patch.
- `stet-commit-f8ae1177` — 1,312 lines, 9 files; judges preferred Opus’s patch.

### Both solved

- `stet-15439c21` — 5 lines, 2 files; Haiku preferred.
- `stet-3169aefa` — 6 lines, 1 file.
- `stet-89dfbc27` — 9 lines, 1 file; Haiku preferred.
- `stet-5276aa3f` — 23 lines, 2 files.
- `stet-6f84e978` — 23 lines, 1 file; Opus preferred.
- `stet-b28f4fe7` — 29 lines, 1 file; Haiku preferred.
- `stet-commit-5d766dde` — 218 lines, 2 files.
- `stet-commit-bbbbae09` — 941 lines, 7 files; Haiku preferred.

### Neither solved

- `stet-commit-e928166f` — 149 lines, 2 files.
- `stet-commit-30506f8e` — 476 lines, 17 files.
- `stet-commit-5e075641` — 481 lines, 3 files.
- `stet-commit-a8716b88` — 564 lines, 10 files.
- `stet-commit-87814583` — 763 lines, 8 files.
- `stet-commit-e41c2e2d` — 977 lines, 5 files.
- `stet-commit-2826765e` — 1,420 lines, 19 files.

Two of 24 tasks are left out because one setup has no scored result on them. A task counts as solved when tests pass or a judge finds the fix equivalent to the merged change; Stet score also weighs code quality.

## Cost and time on each task

The comparison uses a log scale; filled points indicate solved tasks, hollow points indicate unsolved tasks, and the line marks the median.

- **Haiku 5.5 max:** mean $3.46, median $2.46.
- **Opus 4.8 medium:** mean $7.46, median $3.18.
- Haiku cost less on 17 of 22 compared tasks; Opus cost less on 5.
- **Haiku 5.5 max:** mean 58.2 minutes, median 46.7 minutes.
- **Opus 4.8 medium:** mean 23.8 minutes, median 17.3 minutes.
- Opus finished faster on 21 of 22 tasks; Haiku finished faster on 1.

## Where each setup is the better choice

### Small fixes

10 tasks, 2–84 changed lines in 1–2 files. Opus solved 9/10 versus Haiku’s 7/10; Haiku cost 1.1× less and Opus was 3.1× faster.

| Metric | Haiku 5.5 max | Opus 4.8 medium |
|---|---:|---:|
| Solved | 7/10 | 9/10 |
| Cost / task | $1.11 | $1.22 |
| Agent time / task | 36.4 min | 11.6 min |

### Large, cross-cutting changes

14 tasks, 113–1,420 changed lines in up to 19 files. Opus solved 4/12 versus Haiku’s 3/12; Haiku cost 2.3× less and Opus was 2.2× faster.

| Metric | Haiku 5.5 max | Opus 4.8 medium |
|---|---:|---:|
| Solved | 3/12 | 4/12 |
| Cost / task | $5.42 | $12.66 |
| Agent time / task | 76.3 min | 34.0 min |

## Side by side

| Metric | Haiku 5.5 max | Opus 4.8 medium |
|---|---|---|
| Harness | `claude-code` | `claude-code` |
| Reasoning effort | max | medium |
| Rank | 10 of 18 | 13 of 18 |
| Stet score | 48.7, 95% range 38.3–58.7 | 46.8, 95% range 38.0–55.7 |
| Solved | 10/24: 7 passed, 3 judged equivalent | 13/24: 11 passed, 2 judged equivalent |
| Cost / task | $3.83 | $7.68 |
| Agent time / task | 58.4 min | 23.3 min |

## Research links

- [Cost per successful task ranked the coding agent that lost every head to head 1st](https://www.stet.sh/blog/cost-per-successful-task)
- [A benchmark mined from my own repo agreed with FrontierCode](https://www.stet.sh/blog/benchmark-agreed-with-frontier-code)
- [GLM 5.2 vs Composer 2.5 on 50 real Go and Rust PRs](https://www.stet.sh/blog/glm-5-2-passes-tests-fails-review)

## Questions

**Which is better for coding?** No clear quality difference: Haiku won 6, Opus won 6, 10 tied, and 2 were undecided. Haiku was 2.0× cheaper; Opus was 2.5× faster.

**How did we test them?** Both agents worked on the same 24 merged changes from private `Stet-AI/stet`, each from the repository state before the change. Tests and judges graded every patch against the merged change; no public benchmark scores were used. A Stet score is a setup’s average chance of beating another setup on a task from 0 to 100; 50 is even with the field.

## Run this comparison on your repository

Stet replays your merged changes, scores quality above pass/fail, and delivers comparison reports through your agent. [Get started](https://www.stet.sh/private)

## About this data

- 24 merged changes from private `Stet-AI/stet`, replayed from the repository state just before each change.
- Data from November 19, 2026.
- Same leaderboard run of 18 setups: `bmk_d4ba310fdf50db4ccb2092e3`.

Links: [See all 18 setups](https://www.stet.sh/leaderboard) · [How Stet measures](https://www.stet.sh/methodology)

## More comparisons

- [Haiku 5.5 max vs DeepSeek V4.1 Flash (10/17) max](https://www.stet.sh/leaderboard/compare/deepseek-v4-1-flash-10-17-max-vs-haiku-5-5-max)
- [Haiku 5.5 max vs Muse Spark 1.3 high](https://www.stet.sh/leaderboard/compare/haiku-5-5-max-vs-muse-spark-1-3-high)
- [Haiku 5.5 max vs DeepSeek V4.1 Flash max](https://www.stet.sh/leaderboard/compare/deepseek-v4-1-flash-max-vs-haiku-5-5-max)
- [Haiku 5.5 max vs GLM 5.3 Flash high](https://www.stet.sh/leaderboard/compare/glm-5-3-flash-high-vs-haiku-5-5-max)
- [Haiku 5.5 max vs Pareto 26.9 medium](https://www.stet.sh/leaderboard/compare/haiku-5-5-max-vs-pareto-26-9-medium)
- [Opus 4.8 medium vs GLM 5.3 Flash high](https://www.stet.sh/leaderboard/compare/glm-5-3-flash-high-vs-opus-4-8-medium)
- [Opus 4.8 medium vs GPT 5.6 Luna max](https://www.stet.sh/leaderboard/compare/gpt-5-6-luna-max-vs-opus-4-8-medium)
- [Opus 4.8 medium vs Muse Spark 1.3 high](https://www.stet.sh/leaderboard/compare/muse-spark-1-3-high-vs-opus-4-8-medium)
- [Opus 4.8 medium vs GPT 6 Luna max](https://www.stet.sh/leaderboard/compare/gpt-6-luna-max-vs-opus-4-8-medium)
- [Opus 4.8 medium vs Haiku 5.5 low](https://www.stet.sh/leaderboard/compare/haiku-5-5-low-vs-opus-4-8-medium)
- [All 153 comparisons](https://www.stet.sh/leaderboard/compare)

---

[Read this page’s summary](https://www.stet.sh/agents/api/public/v1/page-tldr?url=https%3A%2F%2Fwww.stet.sh%2Fleaderboard%2Fcompare%2Fhaiku-5-5-max-vs-opus-4-8-medium&trace=t_5da45423e484)

## Ask a follow-up about this page

Need a more specific answer? Ask the exact question you have. It returns a fresh, cited answer from Stet’s knowledge base. No account or API key required.

Open this URL with your own question, URL-encoded. Include your question in q. All research inputs are optional; no account or setup.

`GET https://www.stet.sh/agents/agent/faq?q={question}&trace=t_5da45423e484`

[Tool guide and examples](https://www.stet.sh/agents/agent-access) · [OpenAPI](https://www.stet.sh/agents/api/public/v1/openapi.json) · [Page TLDR form](https://www.stet.sh/agents/agent-tools/page-tldr?url=https%3A%2F%2Fwww.stet.sh%2Fleaderboard%2Fcompare%2Fhaiku-5-5-max-vs-opus-4-8-medium). Opening a form does not execute a tool.