Skip to content

Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on Implement npm-style semver range matching (caret/tilde/x/hyphen, prerelease rule) from a written spec

AClaude Haiku 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-17 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.03
Wall time
2m
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Task: implement semver range matching Implement `compare`, `satisfies` and `max_satisfying` in `semrange.py` following `RANGES.md` exactly (it is a subset of npm's node-semver semantics; prerelease handling is graded). Python standard library only; no third-party packages, no network. Add tests if you want.

Desugaring partial versions with implicit -0 upper bounds, caret on 0.x/0.0.x, hyphen ranges with partial ends, and the same-tuple prerelease rule are classic places where implementations diverge from npm. 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).

Success criteria: 1. All 49 hidden unittest cases pass (60 s cap): 46 satisfies() vectors, compare() precedence incl. prerelease ordering and build metadata, ValueError on invalid versions/ranges, max_satisfying() incl. skipping invalid strings and returning the original string.

Setup notes: Use Case Scout batch 5 case e11-semver-ranges (offline fixture). Prepared with batch5/prepare.sh; hidden key/tests kept outside the agent's work dir. Prompt = TASK.md verbatim. Scored with the scout's scorer: `python score.py WORKDIR KEYDIR`. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. Rescored 2026-10-16 with Scout's fixed scorer (see batch5/CHANGES-2026-10-17.md). The original score was 48/49, and the model's output is unchanged. Claude Code's own reported cost for claude-haiku-5.5: $0.031441. Score: 49/49. 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
$0.03
Wall time
2m
Human step-ins
0
Turns
14
Model usage
ModelInput tokOutput tokCost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on634,67333,497$0.03

Reproduce this run

  1. Use the use case prompt on the use case page.
  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: 1. All 49 hidden unittest cases pass (60 s cap): 46 satisfies() vectors, compare() precedence incl. prerelease ordering and build metadata, ValueError on invalid versions/ranges, max_satisfying() incl. skipping invalid strings and returning the original string.
  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": 144581,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.03144065,
  "usage": {
   "input_tokens": 28,
   "cache_creation_input_tokens": 43910,
   "cache_read_input_tokens": 590735,
   "output_tokens": 33497,
   "output_tokens_details": {
    "thinking_tokens": 24611
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 43910,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 710,
     "cache_read_input_tokens": 55786,
     "cache_creation_input_tokens": 663,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 663
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-haiku-5-5": {
    "inputTokens": 28,
    "outputTokens": 33497,
    "cacheReadInputTokens": 590735,
    "cacheCreationInputTokens": 43910,
    "webSearchRequests": 0,
    "costUSD": 0.03144065,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 24611,
    "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": 14,
  "subtype": "success",
  "api_error_status": null,
  "result": "I implemented `compare`, `satisfies` and `max_satisfying` in `semrange.py` following `RANGES.md`. The 21 tests I wrote in `test_semrange.py` pass (`python3 -m unittest test_semrange`). They cover the spec's examples, including the prerelease cases. Standard library only.\n\nThe approach: each version becomes a sortable key tuple, and each range comparator is desugared into plain `>=`, `<`, `<=`, `>`, `=` bounds. Comparators the user wrote with a prerelease are flagged, and the flag decides the prerelease rule. Implicit `-0` upper bounds are not flagged, as the spec requires.\n\nJudgment calls where the spec was silent or ambiguous:\n\n- **Empty `||` alternative** (e.g. `1.0.0 ||`) is treated as `*`, matching npm. Say if you want it to raise `ValueError` instead.\n- **`max_satisfying` raises `ValueError`** if any version string is invalid, rather than skipping it as npm does. This follows the general \"invalid version \u2192 ValueError\" rule.\n- **Strict version syntax:** no surrounding whitespace in versions, leading zeros rejected, and only one `v` or `=` prefix. `~>` is rejected.\n- **Partial versions** with a wildcard followed by a number (`1.x.3`) and prereleases on partials (`1.2-beta`) raise `ValueError`.\n- **Whitespace** is allowed between an operator and its version, and also between `~`/`^` and their version (npm does this too).\n\nNo git repository here, so nothing was committed.",
  "ttft_ms": 1187,
  "type": "result",
  "duration_ms": 145307,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 940,
  "time_to_request_ms": 199,
  "first_content_frame_ms": 940,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly