This scoreboard answers one question: how reliably does a model operate the Cast harness?
It is not a general intelligence ranking. A model is certified once it clears 80%
of the full behavior suite across exactly three fresh attempts per case, where every attempt on a case must
agree (a case that only passes sometimes doesn't count). Core/Chain break
the same score down by single-turn tool contracts vs. multi-turn stateful workflows (see
Behavior Evals).
Tokens are per-attempt averages; the time and turns columns are avg/median/p75/p95/p99 over every
individual attempt. See
Eval Methodology
for the full methodology.
Model
Reasoning
Score
Passed
Certified
Core
Chain
Time (avg/p50/p75/p95/p99)
Turns (avg/p50/p75/p95/p99)
Input tokens (avg/p50/p75/p95/p99)
Output tokens (avg/p50/p75/p95/p99)
Provider URL
Last updated
mimo-v2.6-flash
high
92.9%
39/42
✓ certified
85% (11/13)
97% (28/29)
36.6s / 14.5s / 44.1s / 138.6s / 300.0s
4.3 / 3.0 / 5.0 / 12.0 / 16.0
43.7k / 32.0k / 48.5k / 125.8k / 176.6k
1.0k / 367 / 1.0k / 3.7k / 8.7k
https://token-plan-sgp.xiaomimimo.com/v1
2026-09-25
mimo-v2.6-pro
high
92.9%
39/42
✓ certified
92% (12/13)
93% (27/29)
42.9s / 25.3s / 44.2s / 150.0s / 240.0s
3.8 / 3.0 / 4.0 / 9.0 / 13.0
38.5k / 26.3k / 46.8k / 92.8k / 163.3k
1.2k / 512 / 1.4k / 4.2k / 8.1k
https://token-plan-sgp.xiaomimimo.com/v1
2026-09-25
openai/gpt-5.6-luna
medium
92.1%
35/38
✓ certified
92% (11/12)
92% (24/26)
11.4s / 8.5s / 11.7s / 30.1s / 77.1s
3.5 / 3.0 / 4.0 / 10.0 / 13.0
28.7k / 19.7k / 29.4k / 78.6k / 150.4k
478 / 255 / 485 / 1.4k / 4.6k
https://openrouter.ai/api/v1
2026-08-09
mimo-v2.6-flash
off
88.1%
37/42
✓ certified
77% (10/13)
93% (27/29)
27.2s / 11.6s / 32.9s / 80.3s / 300.0s
4.1 / 3.0 / 4.0 / 12.0 / 14.0
42.6k / 24.7k / 43.6k / 125.1k / 223.3k
467 / 196 / 490 / 1.3k / 3.9k
https://token-plan-sgp.xiaomimimo.com/v1
2026-09-25
deepseek/deepseek-v4-flash-0731
high
86.8%
33/38
✓ certified
92% (11/12)
85% (22/26)
13.2s / 7.7s / 14.1s / 34.3s / 126.1s
3.1 / 3.0 / 4.0 / 6.0 / 9.0
29.4k / 23.6k / 34.0k / 69.6k / 107.7k
717 / 350 / 763 / 1.7k / 8.5k
https://openrouter.ai/api/v1
2026-08-09
deepseek/deepseek-v4-pro
high
84.2%
32/38
✓ certified
100% (12/12)
77% (20/26)
17.3s / 9.6s / 15.3s / 34.4s / 240.0s
3.1 / 3.0 / 4.0 / 7.0 / 10.0
29.4k / 23.1k / 31.1k / 67.7k / 116.5k
873 / 309 / 661 / 1.6k / 16.8k
https://openrouter.ai/api/v1
2026-08-09
mimo-v2.6-pro
off
66.7%
28/42
not yet
77% (10/13)
62% (18/29)
33.7s / 15.3s / 30.8s / 180.0s / 300.0s
4.9 / 3.0 / 5.0 / 14.0 / 28.0
53.7k / 31.9k / 51.6k / 214.8k / 396.5k
735 / 252 / 764 / 2.4k / 9.6k
https://token-plan-sgp.xiaomimimo.com/v1
2026-09-25
Per-signal breakdown
mimo-v2.6-flash — signal breakdown
Signal
Passed
argument-grounding
9/11
background-lifecycle
3/3
delegation
4/4
filesystem-safety
9/9
mcp-discovery
2/2
mcp-tool-chain
1/1
no-unneeded-tools
10/11
parallel-tools
2/2
plan-lifecycle
5/6
plan-safety
1/1
required-tool
6/8
skill-discovery
2/2
state-persistence
8/8
state-transition
1/2
tool-chain
12/12
tool-error-recovery
8/8
tool-result-integrity
8/8
mimo-v2.6-pro — signal breakdown
Signal
Passed
argument-grounding
10/11
background-lifecycle
3/3
delegation
3/4
filesystem-safety
9/9
mcp-discovery
2/2
mcp-tool-chain
1/1
no-unneeded-tools
10/11
parallel-tools
2/2
plan-lifecycle
6/6
plan-safety
1/1
required-tool
7/8
skill-discovery
2/2
state-persistence
7/8
state-transition
1/2
tool-chain
12/12
tool-error-recovery
7/8
tool-result-integrity
8/8
openai/gpt-5.6-luna — signal breakdown
Signal
Passed
argument-grounding
9/10
background-lifecycle
3/3
delegation
3/4
filesystem-safety
9/9
mcp-discovery
2/2
mcp-tool-chain
1/1
no-unneeded-tools
11/11
parallel-tools
1/2
plan-lifecycle
5/6
plan-safety
1/1
required-tool
7/8
skill-discovery
2/2
state-persistence
5/5
state-transition
2/2
tool-chain
10/10
tool-error-recovery
5/6
tool-result-integrity
8/8
mimo-v2.6-flash — signal breakdown
Signal
Passed
argument-grounding
8/11
background-lifecycle
3/3
delegation
3/4
filesystem-safety
9/9
mcp-discovery
2/2
mcp-tool-chain
1/1
no-unneeded-tools
9/11
parallel-tools
2/2
plan-lifecycle
5/6
plan-safety
1/1
required-tool
6/8
skill-discovery
2/2
state-persistence
8/8
state-transition
0/2
tool-chain
12/12
tool-error-recovery
8/8
tool-result-integrity
7/8
deepseek/deepseek-v4-flash-0731 — signal breakdown
