Hollow markers & dashed spokes: axis not yet scored
G
79
A
92.5
U
78.7
N
80
T
100
L
—
E
70
T
7.9
Specification
- Parameters
- 24B
- Architecture
- mistral3
- Size on disk
- 25.93 GB
- Quantization
- 8bit
- Format
- MLX
- Reasoning (CoT)
- No
- Internal ID
- M13
- Mean speed
- 12.9 tok/s across suites
- Stall census
- 0 stalls in 54 observed tests (0.0%)
- Model card
- huggingface.co
Suite results
| Suite | Score | Avg / 20 | Tests | tok/s |
|---|---|---|---|---|
| General capability (13-task real-workload suite) 1a | 205.4 / 260 | 15.8 | 13/13 | 20.3 |
| Agentic tool-calling & protocol adherence 1b | 148 / 160 | 18.5 | 8/8 | 17.1 |
| Coding depth 1c | 84 / 120 | 14 | 6/6 | 14.8 |
| Doc/OCR vision 1d | 144 / 160 | 18 | 8/8 | 14 |
| Doc/OCR — real-degraded tier 1d2 | 44.8 / 80 | 11.2 | 4/4 | 11.4 |
| Content-production depth 1e | 85.2 / 120 | 14.2 | 6/6 | 19.7 |
| Long-context retrieval & synthesis 1g | 117 / 120 | 19.5 | 6/6 | 2.5 |
| Long-context multi-needle (MRCR) 1g2 | 27 / 60 | 9 | 3/3 | 3.6 |
Each test is LLM-judged 0–20 against a fixed rubric; suite max = tests × 20. Where fewer tests ran than the suite total, unrun tests count as zero toward the axis score. See methodology.