Suite results
Suite Score Avg / 20 Tests n tok/s Run
▶ General capability (13-task real-workload suite) 1a 175.5 / 260 13.5 13/13 n=1 60.4 run 2026-09-06
| Test | Category | Score | | tok/s |
| A1 | Dev/Ops Scripting | 8/20 | | 63 |
| A2 | Dev/Ops Scripting | 12/20 | | 60.9 |
| A3 | Dev/Ops Scripting | 9/20 | | 56.9 |
| A4 | Dev/Ops Scripting | 9/20 | | 55.9 |
| A5 | Dev/Ops Scripting | 9/20 | | 61.7 |
| A6 | Dev/Ops Scripting | 20/20 | | 62.8 |
| B1 | Document Processing | 20/20 | | 60.8 |
| B2 | Document Processing | 20/20 | | 61.5 |
| B3 | Document Processing | 9/20 | | 57.9 |
| C1 | Content Production | 18/20 | | 59.8 |
| C2 | Content Production | 16/20 | | 52.1 |
| C3 | Content Production | 16/20 | | 50.3 |
| C4 | Content Production | 9/20 | | 51.3 |
Category mean Dev/Ops Scripting 11.2Document Processing 16.3Content Production 14.8
▶ Agentic tool-calling & protocol adherence 1b 147.2 / 160 18.4 8/8 n=1 66.6 run 2026-09-06
| Test | Category | Score | | tok/s |
| WA1 | Tool Calling & Protocol Adherence | 20/20 | | 44.2 |
| WA2 | Tool Calling & Protocol Adherence | 20/20 | | 48.8 |
| WA3 | Tool Calling & Protocol Adherence | 20/20 | | 50.8 |
| WA4 | Tool Calling & Protocol Adherence | 20/20 | | 45 |
| WA5 | Tool Calling & Protocol Adherence | 20/20 | | 56.5 |
| WB1 | Agent Robustness & State Management | 17/20 | | 62.8 |
| WB2 | Agent Robustness & State Management | 19/20 | | 64.1 |
| WB3 | Agent Robustness & State Management | 11/20 | | 57.2 |
Category mean Tool Calling & Protocol Adherence 20Agent Robustness & State Management 15.7
▶ Coding depth 1c 125.1 / 180 13.9 9/9 n=1 59.8 run 2026-10-02
| Test | Category | Score | | tok/s |
| K1 | Coding — Concurrency | 18/20 | | 62.5 |
| K2 | Coding — Algorithms | 8/20 | | 62.8 |
| K3 | Coding — Security Review | 9/20 | | 62.3 |
| K4 | Coding — Performance | 19/20 | | 62.4 |
| K5 | Coding — Refactoring Under Constraints | 18/20 | | 59.6 |
| K6 | Coding — Test Writing | 17/20 | | 54.5 |
| K7 | Coding — React Component (typed) | 12/20 | | 54.1 |
| K8 | Coding — TypeScript Type System | 13/20 | | 53.6 |
| K9 | Coding — React Debugging | 11/20 | | 54 |
Category mean Coding — Concurrency 18Coding — Algorithms 8Coding — Security Review 9Coding — Performance 19Coding — Refactoring Under Constraints 18Coding — Test Writing 17Coding — React Component (typed) 12Coding — TypeScript Type System 13Coding — React Debugging 11
▶ Doc/OCR vision 1d 92.8 / 160 11.6 8/8 n=1 62.7 run 2026-09-25
| Test | Category | Score | | tok/s |
| D1 | Doc/OCR — Transcription | 7/20 | | 47.2 |
| D2 | Doc/OCR — Document QA | 12/20 | | 40.1 |
| D3 | Doc/OCR — Field Extraction | 16/20 | | 50.7 |
| D4 | Doc/OCR — Document QA | 6/20 | | 45.5 |
| D5 | Doc/OCR — Transcription | 7/20 | | 56.9 |
| D6 | Doc/OCR — Field Extraction | 20/20 | | 54.6 |
| D7 | Doc/OCR — Degraded Input | 10/20 | | 59.3 |
| D8 | Doc/OCR — Degraded Input | 15/20 | | 55.9 |
Category mean Doc/OCR — Transcription 7Doc/OCR — Document QA 9Doc/OCR — Field Extraction 18Doc/OCR — Degraded Input 12.5
▶ Doc/OCR — real-degraded tier 1d2 36.8 / 80 9.2 4/4 n=1 62.3 run 2026-09-22
| Test | Category | Score | | tok/s |
| D9 | Doc/OCR — Real Degraded Transcription | 3/20 | | 47.9 |
| D10 | Doc/OCR — Handwriting Extraction | 17/20 | | 46.7 |
| D11 | Doc/OCR — Degraded Table Transcription | 7/20 | | 52.1 |
| D12 | Doc/OCR — Degraded Table QA | 10/20 | | 38 |
Category mean Doc/OCR — Real Degraded Transcription 3Doc/OCR — Handwriting Extraction 17Doc/OCR — Degraded Table Transcription 7Doc/OCR — Degraded Table QA 10
▶ Content-production depth 1e 93 / 120 15.5 6/6 n=1 54.1 run 2026-09-30
| Test | Category | Score | | tok/s |
| E1 | SOW & Requirements Decomposition | 20/20 | | 58.1 |
| E2 | SOW & Requirements Decomposition | 18/20 | | 58 |
| E3 | Executive & Proposal Content | 17/20 | | 50.6 |
| E4 | Executive & Proposal Content | 5/20 | | 48.9 |
| E5 | Executive & Proposal Content | 18/20 | | 49.7 |
| E6 | Executive & Proposal Content | 15/20 | | 48 |
Category mean SOW & Requirements Decomposition 19Executive & Proposal Content 13.8
▶ Long-context retrieval & synthesis 1g 106.2 / 120 17.7 6/6 n=1 51.4 run 2026-09-25
| Test | Category | Score | | tok/s |
| L1 | Long-Context — Needle Retrieval | 20/20 | | 5.8 |
| L2 | Long-Context — Grounded QA with Distractors | 20/20 | | 2.8 |
| L3 | Long-Context — Ops Log Reasoning | 8/20 | | 5.7 |
| L4 | Long-Context — Faithful Summarization | 20/20 | | 17 |
| L5 | Long-Context — Instruction Retention | 18/20 | | 38.8 |
| L6 | Long-Context — Full-Haystack QA | 20/20 | | 1.6 |
Category mean Long-Context — Needle Retrieval 20Long-Context — Grounded QA with Distractors 20Long-Context — Ops Log Reasoning 8Long-Context — Faithful Summarization 20Long-Context — Instruction Retention 18Long-Context — Full-Haystack QA 20
▶ Long-context multi-needle (MRCR) 1g2 15.9 / 60 5.3 3/3 n=1 42.9 run 2026-09-21
| Test | Category | Score | | tok/s |
| L7 | Long-Context — Multi-Needle Retrieval (MRCR) | 8/20 | | 15.6 |
| L8 | Long-Context — Multi-Needle Retrieval (MRCR) | 4/20 | | 21.7 |
| L9 | Long-Context — Multi-Needle Retrieval (MRCR) | 4/20 | | 7.1 |
▶ Live one-shot builds (runtime-verified) 1h 36 / 80 9 4/4 n=1 55 run 2026-09-22
| Test | Category | Score | | tok/s |
| H1 | Web-Estate One-Shot Builds | 12/20 | | 61.4 |
| H2 | Web-Estate One-Shot Builds | 12/20 | | 57 |
| H3 | Web-Estate One-Shot Builds | 7/20 | | 53.3 |
| H4 | Web-Estate One-Shot Builds | 5/20 | | 48.5 |
▶ Production replay (real agent workload) 1i 115.2 / 240 9.6 12/12 n=1 39.6 run 2026-10-10
| Test | Category | Score | | tok/s |
| I1 | Nightly Fix Replay | 7/20 | | 34.7 |
| I2 | Nightly Fix Replay | 10/20 | | 31.9 |
| I3 | Nightly Fix Replay | 3/20 | | 24.8 |
| I4 | Nightly Fix Replay | 5/20 | | 30.8 |
| I5 | Nightly Fix Replay | 2/20 | | 30.8 |
| I6 | Nightly Fix Replay | 4/20 | | 26.5 |
| I7 | Audit Judgment | 16/20 | | 39.6 |
| I8 | Audit Judgment | 18/20 | | 39.5 |
| I9 | Cron Dependency Reasoning | 10/20 | | 31.8 |
| I10 | Cron Dependency Reasoning | 13/20 | | 29.5 |
| I11 | Curation Judgment | 13/20 | | 35.4 |
| I12 | Curation Judgment | 14/20 | | 37.1 |
Category mean Nightly Fix Replay 5.2Audit Judgment 17Cron Dependency Reasoning 11.5Curation Judgment 13.5
▶ Brand-constrained static build (negative-constraint adherence) 1j 68 / 100 13.6 5/5 n=1 49.2 run 2026-09-24
| Test | Category | Score | | tok/s |
| J1 | Brand-Constrained Static Build | 15/20 | | 45.4 |
| J3 | Brand-Constrained Static Build | 20/20 | | 47.3 |
| J4 | Brand-Constrained Static Build | 10/20 | | 46.3 |
| J5 | Brand-Constrained Static Build | 10/20 | | 49.9 |
| J6 | Brand-Constrained Static Build | 13/20 | | 50.4 |
Each test is LLM-judged 0–20 against a fixed rubric; suite max = tests × 20.
An ADJ chip marks a cell carrying a subjective
adjustment from the maintainer; the judged score is the one shown, is unchanged,
and remains what the axes, ranks and tie bands are computed from. Hover for the reason.
Rows with a ▶ expand to the per-test breakdown.
Where fewer tests ran than the suite total, unrun tests count as zero toward the axis score.
See methodology.