Skip to content

Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on Model ladder: fix 10 real merged bugs with tests hidden

AClaude Haiku 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-16 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.0058
Wall time
19s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt (this run)

You are working in a checkout of the marshmallow-code/marshmallow repository. A user filed this bug report: --- validate.URL rejects a URL whose fragment directly follows the host ```python from marshmallow import validate validate.URL()("http://example.org#fragment") # ValidationError: Not a valid URL. validate.URL()("http://user@example.com#fragment") # same ``` `http://example.org/#fragment` and `http://example.org?a=1#fragment` are accepted, so only the bare-host + fragment case fails. RFC 3986 allows a fragment after an empty path, and links like `https://example.com#section` are common. Relatedly, with `relative=True` (and in relative-only mode) a bare same-document reference such as `#frag` is rejected, although a bare `?query` is accepted. --- Fix the bug in the library code. Dependencies are already installed (Python virtualenv in .venv; run tests with .venv/bin/pytest). Keep the change minimal and in the style of the codebase; you may add or update tests. Make sure the existing test suite still passes. You have no internet access.

Five real bug fixes merged into popular open-source libraries after the models' training cutoff (humanize, marshmallow, tenacity, nanoid, path-to-regexp). Each agent starts from the commit before the fix with only a symptom-style bug report, and the fix PR's own regression tests are applied afterwards to check the result. The size of the agent's source change is compared with the human fix. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked). Upstream fix being reproduced: https://github.com/marshmallow-code/marshmallow/pull/3016, https://github.com/marshmallow-code/marshmallow/commit/902f99c4151d1f0c4c3b8e8dfbafc7b6a76aeaa6.

Success criteria: Hidden PR tests pass; existing suite stays green; diff under 2x the human diff size.

Setup notes: Real merged bug fix: marshmallow-code/marshmallow https://github.com/marshmallow-code/marshmallow/pull/3016 (merged after the models' Jun 2026 cutoff). Agent got a history-free snapshot of the parent commit 3207be3147 with deps installed, and a symptom-only bug report written by the runner from the PR description (no fix hints). The PR's test files were overlaid only after the agent finished. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Diff-size flag (not scored): agent changed 2 source lines vs 2 in the human fix (allowance 20; within allowance). Claude Code's own reported cost for claude-haiku-5.5: $0.005783. Score: 2/2. Cost is an ESTIMATE: measured tokens x pricing.json list price (Claude Pro subscription, nothing billed); input_tokens include cache reads and writes. Single run.

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost
$0.0058
Wall time
19s
Human step-ins
0
Turns
11
Model usage
ModelInput tokOutput tokCost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on138,4613,040$0.0058

Reproduce this run

  1. Use the exact prompt shown above for this run (the use case page has the shared prompt).
  2. Same setup: Claude Haiku 5.5 (anthropic/claude-haiku-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
  3. It counts as a success only if it meets every success criterion: Hidden PR tests pass; existing suite stays green; diff under 2x the human diff size.
  4. Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Evidence log
[
 {
  "duration_api_ms": 15871,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.00578286,
  "usage": {
   "input_tokens": 12,
   "cache_creation_input_tokens": 15143,
   "cache_read_input_tokens": 123306,
   "output_tokens": 3040,
   "output_tokens_details": {
    "thinking_tokens": 621
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 15143,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 460,
     "cache_read_input_tokens": 25830,
     "cache_creation_input_tokens": 1852,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 1852
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-haiku-5-5": {
    "inputTokens": 12,
    "outputTokens": 3040,
    "cacheReadInputTokens": 123306,
    "cacheCreationInputTokens": 15143,
    "webSearchRequests": 0,
    "costUSD": 0.00578286,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 621,
    "canonicalModel": "claude-haiku-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 11,
  "subtype": "success",
  "api_error_status": null,
  "result": "Fixed. `validate.URL` now accepts a fragment directly after the host, and a bare `#frag` in relative mode. The full suite passes: 1192 tests.\n\n**Cause:** In `src/marshmallow/validate.py`, the regex's `relative_part` only allowed the remainder after the host to start with `/` or `?`. A `#` right after the host therefore never matched.\n\n**Change:** `relative_part` now allows `#` as well:\n\n```python\nrelative_part = r\"(?:/?|[/?#]\\S+)\\Z\"\n```\n\nThis covers both reported cases. `http://example.org#fragment` and `http://[REDACTED:email]#fragment` now pass, and `#frag` passes with `relative=True` and with relative-only mode.\n\n**Tests** added to `tests/test_validate.py`:\n- Absolute valid: `http://example.org#fragment`, `http://[REDACTED:email]#fragment`, `https://example.com#section`, `http://example.org/#fragment`, `http://example.org?a=1#fragment`.\n- Relative valid: `#frag` and `http://example.org#fragment`.\n- Relative-only valid: `#frag`.\n\nThe existing invalid-URL cases still reject as before. I haven't committed the change.",
  "ttft_ms": 1644,
  "type": "result",
  "duration_ms": 18067,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1450,
  "time_to_request_ms": 242,
  "first_content_frame_ms": 1450,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly