Public Calibration
HGB Reporting
Kaelum Bench · FMB-KNOW-ACC-HGB-02
This view covers the concrete 60-task suite only. Its provisional 600-point score uses deterministic verification only. It exposes sanitized score, execution, token, and cost aggregates, with no prompts, responses, expected solutions, task IDs, provider generation IDs, keys, rationales, digests, or file paths.
HGB Reporting · 60-task sub-benchmark
Calibration results
Sanitized lifecycle, coverage, score, token, and cost aggregates for FMB-KNOW-ACC-HGB-02. Results are provisional and specific to the recorded model configuration.
- Sub-benchmark
- HGB Reporting · FMB-KNOW-ACC-HGB-02
- Models
- 8 calibration runs
- Coverage
- 60 tasks per model
- Deterministic
- 600 deterministic points
Execution detail
Model runs
Provisional deterministic results use deterministic verification only.
Provisional · Configuration-specific calibration · Single runs; no statistical ranking.
together · high reasoning
nemotron-3-ultra-550b-a55b — open run evidence
nvidia/nemotron-3-ultra-550b-a55b
60 / 60
78.3%
469.5 / 600
59 valid answers
1 invalid JSON response
0 provider failures
72,451
$0.0756
60 / 60
89.4%
536.2 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
240,228
$0.1892
60 / 60
92.4%
554.3 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
166,578
$0.9205
60 / 60
89.3%
535.6 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
74,392
$0.053
60 / 60
90.7%
544.0 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
101,528
$0.2461
60 / 60
90.2%
541.1 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
377,761
$1.10
60 / 60
85.6%
513.6 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
319,978
$2.12
60 / 60
84.5%
507.1 / 600
60 valid answers
0 invalid JSON responses
0 provider failures
214,446
$0.0483