Suite results
Suite Score Avg / 20 Tests n tok/s Run
▶ General capability (13-task real-workload suite) 1a 234 / 260 18 13/13 n=4 54.9 run 2026-09-02
| Test | Category | Score | | tok/s |
| A1 | Dev/Ops Scripting | 14/20 | | 55.8 |
| A2 | Dev/Ops Scripting | 18/20 | | 54.2 |
| A3 | Dev/Ops Scripting | 20/20 | | 53 |
| A4 | Dev/Ops Scripting | 11/20 | | 56.9 |
| A5 | Dev/Ops Scripting | 14/20 | | 51.6 |
| A6 | Dev/Ops Scripting | 20/20 | | 51.3 |
| B1 | Document Processing | 20/20 | | 54.9 |
| B2 | Document Processing | 20/20 | | 54.4 |
| B3 | Document Processing | 20/20 | | 50.2 |
| C1 | Content Production | 20/20 | | 50.9 |
| C2 | Content Production | 17/20 | | 52.6 |
| C3 | Content Production | 20/20 | | 54.8 |
| C4 | Content Production | 20/20 | | 55.6 |
Category mean Dev/Ops Scripting 16.2Document Processing 20Content Production 19.3
▶ Agentic tool-calling & protocol adherence 1b 151.2 / 160 18.9 8/8 n=1 65.2 run 2026-07-28
| Test | Category | Score | | tok/s |
| WA1 | Tool Calling & Protocol Adherence | 20/20 | | 15.6 |
| WA2 | Tool Calling & Protocol Adherence | 20/20 | | 41.7 |
| WA3 | Tool Calling & Protocol Adherence | 20/20 | | 41.2 |
| WA4 | Tool Calling & Protocol Adherence | 20/20 | | 32.7 |
| WA5 | Tool Calling & Protocol Adherence | 20/20 | | 50.7 |
| WB1 | Agent Robustness & State Management | 20/20 | | 60.4 |
| WB2 | Agent Robustness & State Management | 20/20 | | 58 |
| WB3 | Agent Robustness & State Management | 11/20 | | 49.7 |
Category mean Tool Calling & Protocol Adherence 20Agent Robustness & State Management 17
▶ Coding depth 1c 177.3 / 180 19.7 9/9 n=1 52.6 run 2026-09-17
| Test | Category | Score | | tok/s |
| K1 | Coding — Concurrency | 20/20 | | 55.6 |
| K2 | Coding — Algorithms | 20/20 | | 54.3 |
| K3 | Coding — Security Review | 20/20 | | 52.1 |
| K4 | Coding — Performance | 18/20 | | 51.5 |
| K5 | Coding — Refactoring Under Constraints | 20/20 | | 51.6 |
| K6 | Coding — Test Writing | 20/20 | | 48.4 |
| K7 | Coding — React Component (typed) | 19/20 | | 50.5 |
| K8 | Coding — TypeScript Type System | 20/20 | | 48.5 |
| K9 | Coding — React Debugging | 20/20 | | 50.9 |
Category mean Coding — Concurrency 20Coding — Algorithms 20Coding — Security Review 20Coding — Performance 18Coding — Refactoring Under Constraints 20Coding — Test Writing 20Coding — React Component (typed) 19Coding — TypeScript Type System 20Coding — React Debugging 20
▶ Doc/OCR vision 1d 160 / 160 20 8/8 n=1 22.1 run 2026-09-20
| Test | Category | Score | | tok/s |
| D1 | Doc/OCR — Transcription | 20/20 | | 17.3 |
| D2 | Doc/OCR — Document QA | 20/20 | | 19.6 |
| D3 | Doc/OCR — Field Extraction | 20/20 | | 19.4 |
| D4 | Doc/OCR — Document QA | 20/20 | | 21.1 |
| D5 | Doc/OCR — Transcription | 20/20 | | 18 |
| D6 | Doc/OCR — Field Extraction | 20/20 | | 19.5 |
| D7 | Doc/OCR — Degraded Input | 20/20 | | 22.4 |
| D8 | Doc/OCR — Degraded Input | 20/20 | | 13.3 |
Category mean Doc/OCR — Transcription 20Doc/OCR — Document QA 20Doc/OCR — Field Extraction 20Doc/OCR — Degraded Input 20
▶ Content-production depth 1e 114 / 120 19 6/6 n=1 52.4 run 2026-09-16
| Test | Category | Score | | tok/s |
| E1 | SOW & Requirements Decomposition | 20/20 | | 49.2 |
| E2 | SOW & Requirements Decomposition | 18/20 | | 52.2 |
| E3 | Executive & Proposal Content | 18/20 | | 52.5 |
| E4 | Executive & Proposal Content | 18/20 | | 52.8 |
| E5 | Executive & Proposal Content | 20/20 | | 51.8 |
| E6 | Executive & Proposal Content | 20/20 | | 49.6 |
Category mean SOW & Requirements Decomposition 19Executive & Proposal Content 19
▶ Long-context retrieval & synthesis 1g 120 / 120 20 6/6 n=1 52.3 run 2026-08-06
| Test | Category | Score | | tok/s |
| L1 | Long-Context — Needle Retrieval | 20/20 | | 53.5 |
| L2 | Long-Context — Grounded QA with Distractors | 20/20 | | 41.6 |
| L3 | Long-Context — Ops Log Reasoning | 20/20 | | 33.5 |
| L4 | Long-Context — Faithful Summarization | 20/20 | | 36.9 |
| L5 | Long-Context — Instruction Retention | 20/20 | | 31.9 |
| L6 | Long-Context — Full-Haystack QA | 20/20 | | 17.8 |
Category mean Long-Context — Needle Retrieval 20Long-Context — Grounded QA with Distractors 20Long-Context — Ops Log Reasoning 20Long-Context — Faithful Summarization 20Long-Context — Instruction Retention 20Long-Context — Full-Haystack QA 20
▶ Long-context multi-needle (MRCR) 1g2 44.1 / 60 14.7 3/3 n=1 38.2 run 2026-08-14
| Test | Category | Score | | tok/s |
| L7 | Long-Context — Multi-Needle Retrieval (MRCR) | 20/20 | | 31.1 |
| L8 | Long-Context — Multi-Needle Retrieval (MRCR) | 20/20 | | 17.1 |
| L9 | Long-Context — Multi-Needle Retrieval (MRCR) | 4/20 | | 24.7 |
▶ Live one-shot builds (runtime-verified) 1h 66 / 80 16.5 4/4 n=1 57.9 run 2026-09-03
| Test | Category | Score | | tok/s |
| H1 | Web-Estate One-Shot Builds | 19/20 | | 60.8 |
| H2 | Web-Estate One-Shot Builds | 16/20 | | 58.7 |
| H3 | Web-Estate One-Shot Builds | 14/20 | | 53.9 |
| H4 | Web-Estate One-Shot Builds | 17/20 | | 53.8 |
▶ Brand-constrained static build (negative-constraint adherence) 1j 86 / 100 17.2 5/5 n=1 54.2 run 2026-09-16
| Test | Category | Score | | tok/s |
| J1 | Brand-Constrained Static Build | 19/20 | | 57 |
| J3 | Brand-Constrained Static Build | 17/20 | | 56.8 |
| J4 | Brand-Constrained Static Build | 20/20 | | 50.9 |
| J5 | Brand-Constrained Static Build | 17/20 | | 51.5 |
| J6 | Brand-Constrained Static Build | 13/20 | | 49.1 |
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.