Modelpublic

TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill

sourceHugging Faceapache-2.0updated 8mo agoView on Hugging Face

54likes206downloads

Model Card

GLM 4.7 Flash x Claude 4.5 Opus (High Reasoning)

This model was trained on a small reasoning dataset of Claude Opus 4.5, with reasoning effort set to High.

—🧬 Datasets:
—TeichAI/claude-4.5-opus-high-reasoning-250x

—🏗 Base Model:
—unsloth/GLM-4.7-Flash

—⚡ Use cases:
—Coding
—Science
—Deep Research

—∑ Stats (Dataset)
—Costs: $ 52.30 (USD)
—Total tokens (input + output): 2.13 M

How to run

For specific instructions/commands to serve this model locally using vLLM, SGLang, or transformers please see the instructions from the original model's card

For detailed instructions getting started with Llama.cpp please refer to the unsloth guide

Sampling Parameters

z-ai recommends the following sampling parameters for this model:

Default Settings (Most Tasks)	Terminal Bench, SWE Bench Verified
temperature = 1.0	temperature = 0.7
top_p = 0.95	top_p = 1.0
repeat penalty = disabled or 1.0	repeat penalty = disabled or 1.0

—For general use-case: --temp 1.0 --top-p 0.95
—For tool-calling: --temp 0.7 --top-p 1.0
—If using llama.cpp, set --min-p 0.01 as llama.cpp's default is 0.05
—Sometimes you'll need to experiment what numbers work best for your use-case.

If you experience any issues with these parameters, some users have reported better results when lowering temperature to 0.5-0.6

Benchmarks

[image]

Model Comparison vs Base

[image]

—Base model: zai-org/GLM-4.7-Flash

Benchmark	Base Score	Distilled Score	Delta	Delta %
arc_challenge	0.224403	0.217577	-0.00682594	-0.0304183
gpqadiamondzeroshot	0.262626	0.292929	0.030303	0.115385
hellaswag	0.257817	0.256722	-0.0010954	-0.00424874
ifeval	0.109057	0.112754	0.00369686	0.0338983
mmlu	0.229454	0.240706	0.011252	0.0490379
truthfulqa_mc2	0.467552	0.466805	-0.000747457	-0.00159866
winogrande	0.468824	0.504341	0.035517	0.0757576

Aggregate Comparison

Benchmarks Compared	Wins vs Base	Ties vs Base	Losses vs Base	Avg Delta
7	4	0	3	0.0103

Detailed Results

Model	Benchmark	Score	Total Questions	Total Correct
zai-org/GLM-4.7-Flash	winogrande	0.468824	1267	594
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	winogrande	0.504341	1267	639
zai-org/GLM-4.7-Flash	arc_challenge	0.224403	1172	263
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	arc_challenge	0.217577	1172	255
zai-org/GLM-4.7-Flash	hellaswag	0.257817	10042	2589
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	hellaswag	0.256722	10042	2578
zai-org/GLM-4.7-Flash	truthfulqa_mc2	0.467552	817	381
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	truthfulqa_mc2	0.466805	817	381
zai-org/GLM-4.7-Flash	mmlu	0.229454	14042	3222
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	mmlu	0.240706	14042	3380
zai-org/GLM-4.7-Flash	ifeval	0.109057	541	59
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	ifeval	0.112754	541	61
zai-org/GLM-4.7-Flash	gpqadiamondzeroshot	0.262626	198	52
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	gpqadiamondzeroshot	0.292929	198	58

MMLU Subject Breakdown

[image]

MMLU Detailed Results

Model	Subject	Benchmark	Score	Total Questions	Total Correct
zai-org/GLM-4.7-Flash	formal_logic	mmluformallogic	0.285714	126	36
zai-org/GLM-4.7-Flash	highschooleuropean_history	mmluhighschooleuropeanhistory	0.218182	165	36
zai-org/GLM-4.7-Flash	highschoolus_history	mmluhighschoolushistory	0.25	204	51
zai-org/GLM-4.7-Flash	highschoolworld_history	mmluhighschoolworldhistory	0.270042	237	63
zai-org/GLM-4.7-Flash	international_law	mmluinternationallaw	0.239669	121	29
zai-org/GLM-4.7-Flash	jurisprudence	mmlu_jurisprudence	0.259259	108	28
zai-org/GLM-4.7-Flash	logical_fallacies	mmlulogicalfallacies	0.220859	163	36
zai-org/GLM-4.7-Flash	moral_disputes	mmlumoraldisputes	0.248555	346	86
zai-org/GLM-4.7-Flash	moral_scenarios	mmlumoralscenarios	0.237989	895	213
zai-org/GLM-4.7-Flash	philosophy	mmlu_philosophy	0.186495	311	58
zai-org/GLM-4.7-Flash	prehistory	mmlu_prehistory	0.216049	324	70
zai-org/GLM-4.7-Flash	professional_law	mmluprofessionallaw	0.245763	1534	377
zai-org/GLM-4.7-Flash	world_religions	mmluworldreligions	0.321637	171	55
zai-org/GLM-4.7-Flash	business_ethics	mmlubusinessethics	0.3	100	30
zai-org/GLM-4.7-Flash	clinical_knowledge	mmluclinicalknowledge	0.215094	265	57
zai-org/GLM-4.7-Flash	college_medicine	mmlucollegemedicine	0.208092	173	36
zai-org/GLM-4.7-Flash	global_facts	mmluglobalfacts	0.18	100	18
zai-org/GLM-4.7-Flash	human_aging	mmluhumanaging	0.313901	223	70
zai-org/GLM-4.7-Flash	management	mmlu_management	0.174757	103	18
zai-org/GLM-4.7-Flash	marketing	mmlu_marketing	0.290598	234	68
zai-org/GLM-4.7-Flash	medical_genetics	mmlumedicalgenetics	0.3	100	30
zai-org/GLM-4.7-Flash	miscellaneous	mmlu_miscellaneous	0.237548	783	186
zai-org/GLM-4.7-Flash	nutrition	mmlu_nutrition	0.22549	306	69
zai-org/GLM-4.7-Flash	professional_accounting	mmluprofessionalaccounting	0.234043	282	66
zai-org/GLM-4.7-Flash	professional_medicine	mmluprofessionalmedicine	0.183824	272	50
zai-org/GLM-4.7-Flash	virology	mmlu_virology	0.283133	166	47
zai-org/GLM-4.7-Flash	econometrics	mmlu_econometrics	0.236842	114	27
zai-org/GLM-4.7-Flash	highschoolgeography	mmluhighschool_geography	0.176768	198	35
zai-org/GLM-4.7-Flash	highschoolgovernmentandpolitics	mmluhighschoolgovernmentand_politics	0.196891	193	38
zai-org/GLM-4.7-Flash	highschoolmacroeconomics	mmluhighschool_macroeconomics	0.202564	390	79
zai-org/GLM-4.7-Flash	highschoolmicroeconomics	mmluhighschool_microeconomics	0.214286	238	51
zai-org/GLM-4.7-Flash	highschoolpsychology	mmluhighschool_psychology	0.192661	545	105
zai-org/GLM-4.7-Flash	human_sexuality	mmluhumansexuality	0.259542	131	34
zai-org/GLM-4.7-Flash	professional_psychology	mmluprofessionalpsychology	0.25	612	153
zai-org/GLM-4.7-Flash	public_relations	mmlupublicrelations	0.218182	110	24
zai-org/GLM-4.7-Flash	security_studies	mmlusecuritystudies	0.187755	245	46
zai-org/GLM-4.7-Flash	sociology	mmlu_sociology	0.238806	201	48
zai-org/GLM-4.7-Flash	usforeignpolicy	mmluusforeign_policy	0.28	100	28
zai-org/GLM-4.7-Flash	abstract_algebra	mmluabstractalgebra	0.22	100	22
zai-org/GLM-4.7-Flash	anatomy	mmlu_anatomy	0.185185	135	25
zai-org/GLM-4.7-Flash	astronomy	mmlu_astronomy	0.177632	152	27
zai-org/GLM-4.7-Flash	college_biology	mmlucollegebiology	0.256944	144	37
zai-org/GLM-4.7-Flash	college_chemistry	mmlucollegechemistry	0.2	100	20
zai-org/GLM-4.7-Flash	collegecomputerscience	mmlucollegecomputer_science	0.26	100	26
zai-org/GLM-4.7-Flash	college_mathematics	mmlucollegemathematics	0.21	100	21
zai-org/GLM-4.7-Flash	college_physics	mmlucollegephysics	0.215686	102	22
zai-org/GLM-4.7-Flash	computer_security	mmlucomputersecurity	0.28	100	28
zai-org/GLM-4.7-Flash	conceptual_physics	mmluconceptualphysics	0.26383	235	62
zai-org/GLM-4.7-Flash	electrical_engineering	mmluelectricalengineering	0.241379	145	35
zai-org/GLM-4.7-Flash	elementary_mathematics	mmluelementarymathematics	0.208995	378	79
zai-org/GLM-4.7-Flash	highschoolbiology	mmluhighschool_biology	0.174194	310	54
zai-org/GLM-4.7-Flash	highschoolchemistry	mmluhighschool_chemistry	0.152709	203	31
zai-org/GLM-4.7-Flash	highschoolcomputer_science	mmluhighschoolcomputerscience	0.25	100	25
zai-org/GLM-4.7-Flash	highschoolmathematics	mmluhighschool_mathematics	0.211111	270	57
zai-org/GLM-4.7-Flash	highschoolphysics	mmluhighschool_physics	0.198675	151	29
zai-org/GLM-4.7-Flash	highschoolstatistics	mmluhighschool_statistics	0.152778	216	33
zai-org/GLM-4.7-Flash	machine_learning	mmlumachinelearning	0.321429	112	36
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	formal_logic	mmluformallogic	0.206349	126	26
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschooleuropean_history	mmluhighschooleuropeanhistory	0.206061	165	34
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolus_history	mmluhighschoolushistory	0.245098	204	50
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolworld_history	mmluhighschoolworldhistory	0.270042	237	63
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	international_law	mmluinternationallaw	0.239669	121	29
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	jurisprudence	mmlu_jurisprudence	0.305556	108	33
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	logical_fallacies	mmlulogicalfallacies	0.214724	163	35
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	moral_disputes	mmlumoraldisputes	0.271676	346	93
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	moral_scenarios	mmlumoralscenarios	0.222346	895	199
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	philosophy	mmlu_philosophy	0.228296	311	71
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	prehistory	mmlu_prehistory	0.271605	324	88
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	professional_law	mmluprofessionallaw	0.252934	1534	388
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	world_religions	mmluworldreligions	0.280702	171	48
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	business_ethics	mmlubusinessethics	0.3	100	30
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	clinical_knowledge	mmluclinicalknowledge	0.267925	265	71
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	college_medicine	mmlucollegemedicine	0.213873	173	37
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	global_facts	mmluglobalfacts	0.32	100	32
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	human_aging	mmluhumanaging	0.327354	223	73
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	management	mmlu_management	0.213592	103	22
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	marketing	mmlu_marketing	0.286325	234	67
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	medical_genetics	mmlumedicalgenetics	0.35	100	35
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	miscellaneous	mmlu_miscellaneous	0.254151	783	199
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	nutrition	mmlu_nutrition	0.222222	306	68
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	professional_accounting	mmluprofessionalaccounting	0.244681	282	69
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	professional_medicine	mmluprofessionalmedicine	0.183824	272	50
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	virology	mmlu_virology	0.325301	166	54
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	econometrics	mmlu_econometrics	0.280702	114	32
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolgeography	mmluhighschool_geography	0.207071	198	41
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolgovernmentandpolitics	mmluhighschoolgovernmentand_politics	0.176166	193	34
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolmacroeconomics	mmluhighschool_macroeconomics	0.217949	390	85
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolmicroeconomics	mmluhighschool_microeconomics	0.222689	238	53
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolpsychology	mmluhighschool_psychology	0.209174	545	114
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	human_sexuality	mmluhumansexuality	0.21374	131	28
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	professional_psychology	mmluprofessionalpsychology	0.259804	612	159
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	public_relations	mmlupublicrelations	0.309091	110	34
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	security_studies	mmlusecuritystudies	0.159184	245	39
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	sociology	mmlu_sociology	0.253731	201	51
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	usforeignpolicy	mmluusforeign_policy	0.25	100	25
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	abstract_algebra	mmluabstractalgebra	0.23	100	23
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	anatomy	mmlu_anatomy	0.251852	135	34
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	astronomy	mmlu_astronomy	0.164474	152	25
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	college_biology	mmlucollegebiology	0.263889	144	38
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	college_chemistry	mmlucollegechemistry	0.22	100	22
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	collegecomputerscience	mmlucollegecomputer_science	0.22	100	22
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	college_mathematics	mmlucollegemathematics	0.25	100	25
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	college_physics	mmlucollegephysics	0.245098	102	25
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	computer_security	mmlucomputersecurity	0.24	100	24
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	conceptual_physics	mmluconceptualphysics	0.340426	235	80
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	electrical_engineering	mmluelectricalengineering	0.193103	145	28
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	elementary_mathematics	mmluelementarymathematics	0.240741	378	91
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolbiology	mmluhighschool_biology	0.190323	310	58
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolchemistry	mmluhighschool_chemistry	0.216749	203	44
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolcomputer_science	mmluhighschoolcomputerscience	0.19	100	19
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolmathematics	mmluhighschool_mathematics	0.240741	270	65
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolphysics	mmluhighschool_physics	0.172185	151	26
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	highschoolstatistics	mmluhighschool_statistics	0.194444	216	42
TeichAI/GLM-4.7-Flash-Claude-Opus-4.5-High-Reasoning-Distill	machine_learning	mmlumachinelearning	0.241071	112	27

Benchmark Config

—Quantization: 4bit
—Temperature: 0.0
—Top P: 1.0
—Top K: 0
—Repetition Penalty: 1.0

All results were obtained through the official lm evaluation harness

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.