Suite results
Suite Score Avg / 20 Tests n tok/s Run
▶ General capability (13-task real-workload suite) 1a 169.2 / 260 18.8 9/13 ● n=1 78.4 run 2026-09-02
| Test | Category | Score | | tok/s |
| A1 | Dev/Ops Scripting | 20/20 | | 49.9 |
| A2 | Dev/Ops Scripting | 18/20 | | 35.7 |
| A3 | Dev/Ops Scripting | 18/20 | | 52.8 |
| A4 | Dev/Ops Scripting | 20/20 | | 21 |
| A5 | Dev/Ops Scripting | 14/20 | | 53.9 |
| A6 | Dev/Ops Scripting | 20/20 | | 56.2 |
| B1 | Document Processing | 20/20 | | 45.1 |
| B2 | Document Processing | 20/20 | | 23.8 |
| B3 | Document Processing | 20/20 | | 21.3 |
Category mean Dev/Ops Scripting 18.3Document Processing 20
▶ Agentic tool-calling & protocol adherence 1b 148.8 / 160 18.6 8/8 n=1 55.9 run 2026-09-16
| Test | Category | Score | | tok/s |
| WA1 | Tool Calling & Protocol Adherence | 19/20 | | 12.7 |
| WA2 | Tool Calling & Protocol Adherence | 20/20 | | 13.1 |
| WA3 | Tool Calling & Protocol Adherence | 20/20 | | 17.8 |
| WA4 | Tool Calling & Protocol Adherence | 19/20 | | 16.4 |
| WA5 | Tool Calling & Protocol Adherence | 20/20 | | 38.3 |
| WB1 | Agent Robustness & State Management | 20/20 | | 44.6 |
| WB2 | Agent Robustness & State Management | 20/20 | | 36.1 |
| WB3 | Agent Robustness & State Management | 11/20 | | 11.1 |
Category mean Tool Calling & Protocol Adherence 19.6Agent Robustness & State Management 17
▶ Coding depth 1c 162.9 / 180 18.1 9/9 n=1 54.7 run 2026-09-15
| Test | Category | Score | | tok/s |
| K1 | Coding — Concurrency | 20/20 | | 42.8 |
| K2 | Coding — Algorithms | 18/20 | | 48.2 |
| K3 | Coding — Security Review | 16/20 | | 34.7 |
| K4 | Coding — Performance | 19/20 | | 47 |
| K5 | Coding — Refactoring Under Constraints | 20/20 | | 52.7 |
| K6 | Coding — Test Writing | 20/20 | | 40.4 |
| K7 | Coding — React Component (typed) | 11/20 | | 33.3 |
| K8 | Coding — TypeScript Type System | 19/20 | | 26.6 |
| K9 | Coding — React Debugging | 20/20 | | 25.3 |
Category mean Coding — Concurrency 20Coding — Algorithms 18Coding — Security Review 16Coding — Performance 19Coding — Refactoring Under Constraints 20Coding — Test Writing 20Coding — React Component (typed) 11Coding — TypeScript Type System 19Coding — React Debugging 20
▶ Content-production depth 1e 112.8 / 120 18.8 6/6 n=1 82.9 run 2026-08-02
| Test | Category | Score | | tok/s |
| E1 | SOW & Requirements Decomposition | 19/20 | | 53.5 |
| E2 | SOW & Requirements Decomposition | 20/20 | | 35.9 |
| E3 | Executive & Proposal Content | 18/20 | | 71.4 |
| E4 | Executive & Proposal Content | 17/20 | | 64.4 |
| E5 | Executive & Proposal Content | 19/20 | | 55.5 |
| E6 | Executive & Proposal Content | 20/20 | | 29.4 |
Category mean SOW & Requirements Decomposition 19.5Executive & Proposal Content 18.5
▶ Long-context retrieval & synthesis 1g 120 / 120 20 6/6 n=1 64.1 run 2026-09-10
| Test | Category | Score | | tok/s |
| L1 | Long-Context — Needle Retrieval | 20/20 | | 2.5 |
| L2 | Long-Context — Grounded QA with Distractors | 20/20 | | 1.7 |
| L3 | Long-Context — Ops Log Reasoning | 20/20 | | 5.1 |
| L4 | Long-Context — Faithful Summarization | 20/20 | | 12.7 |
| L5 | Long-Context — Instruction Retention | 20/20 | | 3.2 |
| L6 | Long-Context — Full-Haystack QA | 20/20 | | 1.2 |
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 36 / 60 12 3/3 n=1 47.2 run 2026-08-31
| Test | Category | Score | | tok/s |
| L7 | Long-Context — Multi-Needle Retrieval (MRCR) | 20/20 | | 12.8 |
| L8 | Long-Context — Multi-Needle Retrieval (MRCR) | 13/20 | | 8.5 |
| L9 | Long-Context — Multi-Needle Retrieval (MRCR) | 3/20 | | 2.3 |
▶ Live one-shot builds (runtime-verified) 1h 68 / 80 17 4/4 n=1 73 run 2026-09-02
| Test | Category | Score | | tok/s |
| H1 | Web-Estate One-Shot Builds | 19/20 | | 61.8 |
| H2 | Web-Estate One-Shot Builds | 16/20 | | 43 |
| H3 | Web-Estate One-Shot Builds | 17/20 | | 63.9 |
| H4 | Web-Estate One-Shot Builds | 16/20 | | 27.7 |
▶ Brand-constrained static build (negative-constraint adherence) 1j 85 / 100 17 5/5 n=1 55.8 run 2026-09-16
| Test | Category | Score | | tok/s |
| J1 | Brand-Constrained Static Build | 18/20 | | 25.9 |
| J3 | Brand-Constrained Static Build | 20/20 | | 33.5 |
| J4 | Brand-Constrained Static Build | 20/20 | | 25.1 |
| J5 | Brand-Constrained Static Build | 14/20 | | 19.3 |
| J6 | Brand-Constrained Static Build | 13/20 | | 23 |
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.