claude-code + claude-sonnet-5
claude-code harness · Claude Max 20x ($200/mo)
Clears tier 3 of 5 (partial at tiers 4 and 5) on Real-world issues
Capability mean across 2 graded suite families.
What the tiers mean
"Clears" marks the highest rung where every task at it and below was solved. Tier 1: localized single-file bug · tier 3: cause in a different module than the symptom · tier 5: problems that took a human hours. A cleared rung solved every task on it; a partial rung solved some. The headline stops at the last unbroken rung, so a solved tier above a half-solved one counts only as partial.
What the levels mean
Level 1: a full brief naming the surface and the acceptance · level 3: the report as filed · level 5: a vague vibe report in a non-technical user's words. Context falls as the level rises, so vaguer is harder. A cleared level fixed every bug at it and below; a partial level fixed some. The headline stops at the last unbroken level, so a fix above a half-cleared level counts only as partial.
5 of this suite's tasks were credited rather than run: on a suite that asks the same problem at several levels of detail, solving it from the vaguest description credits the more detailed ones instead of asking again.
Per-suite results
11 of 15 tasks solved - grouped by tier, easiest first; hover a square for its time and turns.
Run on claude-code 2.1.226 (Claude Code), model claude-sonnet-5 (--effort medium), week 2026-W33.
One run of this suite consumes about 9.7% of Claude Max 20x's 5h window and 1.0% of the weekly limit.
As work per window: about 1503 suite runs per month on this plan - 7.52 runs per plan-dollar (at $200/mo, priced 2026-08-05). Work figures compare outputs and are comparable across agents; the raw quota share above is not.
How the plan cost was measured
Measured per suite run (reported-delta); an absolute cost on this plan, not comparable with another agent's plan. Estimated: 2 of 15 cells measured no share of their own (the provider returned no reading) and were apportioned one from the harness's own cost telemetry. Measured in a declared quiet window on a shared subscription: operator Claude Max 20x account on eb-runner, shared with the operator IDE panes; quiet-window until a dedicated benchmark login exists.
15 of 25 tasks solved (5 credited without running) - each row is one bug, vaguest report first and context growing to the right; hover a square for its time and turns.
Run on claude-code 2.1.226 (Claude Code), model claude-sonnet-5 (--effort medium), week 2026-W33.
One good solve on this suite consumes, on average, 0.4% of Claude Max 20x's 5h window and 0.2% of the weekly limit.
As work per window: about 36525 good solves per month on this plan - 182.63 solves per plan-dollar (at $200/mo, priced 2026-08-05). Work figures compare outputs and are comparable across agents; the raw quota share above is not.
How the plan cost was measured
Counted over each issue's first solve (5 solves); the walk's failed harder cells and confirmation runs are the benchmark's own search cost and are excluded (full walk: 10.0% over 20 cells). Measured per cell run (reported-delta); an absolute cost on this plan, not comparable with another agent's plan. Measured in a declared quiet window on a shared subscription: operator Claude Max 20x account on eb-runner, shared with the operator IDE panes; quiet-window until a dedicated benchmark login exists.
Task outputs from these runs are withheld: publishing them would reveal the private suite. Per-task scores, times and turns are in the chips above.
Trajectory
One point per grading week - a re-run within a week shows only its latest execution. Grades compare only within a suite version, so each version gets its own line; while a suite is being retired both versions run, and the superseded one is dashed. The filled point is the run the scorecard shows today. Click a point or week for plan usage. All agents over time →
Weekly heads
Newest week first. Click a row for usage.
| Week | Grade | Capability | Usage |
|---|---|---|---|
| 2026-W33 · v3now | D | 53% | 9.7% |
| 2026-W32 · v3 | F | 33% | 8.8% |
Frequently asked questions
Is claude-sonnet-5 worth it on Claude Max 20x?
On our private real-world suite (real-world@v3), claude-sonnet-5 earned grade D with 53% capability, clearing every problem up to tier 3 of 5 - a problem whose cause sits in a different module than the symptom and landing some tier 5 problems. It runs on Claude Max 20x at $200/month (public price as of 2026-08-05). One full suite run consumed about 9.7% of the plan's 5h window. Measured 2026-08-11.
How good is claude-sonnet-5 at real-world coding?
claude-sonnet-5 clears tier 3 of our five-tier real-world ladder - every task at that tier and below - where tier 1 is a localized single-file bug and tier 5 is a problem that took a human engineer hours. It solves some but not all tasks up at tier 5. Its capability score on real-world@v3 is 53% (grade D). Last benchmarked 2026-08-11.
How much coding work does Claude Max 20x buy with claude-sonnet-5?
Roughly 1503 suite runs per month on Claude Max 20x, or 7.52 runs per plan-dollar at $200/month (priced 2026-08-05). Usage is read from the provider's own meter per suite run; the raw quota share is an absolute figure on this plan, not comparable with another agent's plan.
Compare
- claude-sonnet-5 vs gemini-3.1-pro
- claude-sonnet-5 vs gemini-3.6-flash
- claude-sonnet-5 vs claude-fable-5
- claude-sonnet-5 vs claude-haiku-4-5
- claude-sonnet-5 vs claude-opus-5
- claude-sonnet-5 vs gpt-5.6-luna
- claude-sonnet-5 vs gpt-5.6-sol
- claude-sonnet-5 vs gpt-5.6-terra
- claude-sonnet-5 vs composer-2.5
- claude-sonnet-5 vs grok-4.5
In the tools directory
- Claude Code - agentic-coding