GLM-4.7-Flash

+ contribute

GLM-4.7-Flash is an open-weights language model from Z.ai, released January 2026. Its published size is 31.2B parameters (3B active per token) and a 200k-token context window. Licence: mit.

58.8±1.3#208 of 719 LLMsfrom 10 benchmark results across 6 sources

$ how it compares

The LLMs ranked either side of GLM-4.7-Flash, out of 828 with a score, then more from Z.ai.

#	model	score	where	RAM	released
248	Gemma 2 9B IT DPO (Princeton) Princeton NLP	59±1	local	~7.6GB	2024-07
249	Trinity-Large-Preview Arcee AI	59±2	local	~155GB	2026-01
250	GLM-4.7-Flash Z.ai	59±1	api $0.0000/M +local	~19GB	2026-01
251	EXAONE 4.0 32B LG AI Research	59±1	local	~20GB	2025-07
252	Phi-4-reasoning-plus 14B Microsoft	59±1	local	~10GB	2025-04
	more from Z.ai
33	GLM-5.3 Z.ai	76±1	api $2.15/M +local	~239GB	2026-08
52	GLM-5.3-Flash Z.ai	74±1	api $0.24/M +local	~103GB	2026-08
57	GLM-5.2 Z.ai	74±1	api $2.15/M +local	~121GB	2026-06

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasksnot ranked on this task yet · 3 tests

test	result	vs the others in its run	difficulty	counts	why
SWE-bench Verified lab	59.2% ✎	▁▁▁▁▁▁▂▄█▆▁▁ beats 52/131	hard · avg 19%	—	not counted yet: no other model in the same run to compare it with
BrowseComp lab	42.8% ✎	▂·▄▁▂▃▆▄█▂▃▄ beats 15/42	very hard · avg 10%	—	not counted yet: no other model in the same run to compare it with
τ²-bench aggregate lab	79.5% ✎	▁▁▁·▂▂▄▃▃▅█▂ beats 22/37	medium · avg 40%	—	not counted yet: no other model in the same run to compare it with

coding▁▁▁▃▆██▇▆▄▃▁ #339 of 458 * · 12% of its score · 3 tests

Trails Phi-3-mini-128k-instruct (Apr 2024), DeepSeek LLM 67B Chat, WizardLM 70B V1.0 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Luau Leetcode fully solved (Torpedo Software)	14.9% ✎	▁▃▁▁·▁▂▃▅▃█▇ beats 1/49	hard · avg 18%	▇▇▇▇▇ 12%	at its level
SWE-bench Verified lab	59.2% ✎	▁▁▁▁▁▁▂▄█▆▁▁ beats 52/131	hard · avg 19%	—	not counted yet: no other model in the same run to compare it with
LiveCodeBench v6 (window unspecified) lab	64.0% ✎	▂▃▄▄▅██▅▄▇▃▆ beats 28/65	medium · avg 39%	—	not counted yet: no other model in the same run to compare it with

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #142 of 393 * · 15% of its score · 1 test

Beats INTELLECT-3, Nemotron 3 Super 120B-A12B, MiniMax-M2 here, though they rank higher overall. Trails LFM2.5 1.2B Thinking, LFM2.5 8B-A1B, LFM2.5 2.6B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1364.9 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 209/379	avg 1268	▇▇▇▇▇▇ 15%	blunt

math▁▂▃▅▆▅▄█▄▁▁▁ #160 of 398 * · 14% of its score · 2 tests

Beats MiMo-V2.5 (310B-A15B), MiniMax-M2.7, MiniMax-M2.1 here, though they rank higher overall. Trails Olmo 3 7B Think, Qwen3-VL 4B Thinking, Granite 4.2 3B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
OTIS mock AIME 2024–2025	58.3% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 88/193	very hard · avg 8.5%	▇▇▇▇▇▇ 14%	sharp · at its level · noisy
AIME 2025 lab	91.6% ✎	▁·▁▁▁▂▅▆█▇▅▁ beats 118/133	medium · avg 42%	—	not counted yet: no other model in the same run to compare it with

web searchnot ranked on this task yet · 1 test

test	result	vs the others in its run	difficulty	counts	why
BrowseComp lab	42.8% ✎	▂·▄▁▂▃▆▄█▂▃▄ beats 15/42	very hard · avg 10%	—	not counted yet: no other model in the same run to compare it with

other tests9 not in a task · 59% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA Diamond	60.5% ✎	▁▁▁▁▂▄▆▄▅▇█▅ beats 88/208	hard · avg 40%	▇▇▇▇▇▇▇▇▇▇▇ 27%	sharp · at its level
EQ-Bench 3 (emotional intelligence, judge Elo)	1151.3 ✎	▁·▂▁▄▅▅▃█▄▁▁ beats 20/62	avg 1044	▇▇▇▇ 8.9%	typical
EuroEval Life in the UK (English knowledge MC), validation split, 5-shot, accuracy	94.1% ✎	▂·▂▂·▂·▃▃·█▃ beats 13/14	easy · avg 80%	▇▇▇ 8.1%	typical
RobloxQA v2.0 multiple choice (Torpedo Software)	83.1% ✎	▁▁▃▁█▄▃▃▆▂▂▂ beats 3/49	easy · avg 78%	▇▇▇ 7.4%	typical
EuroEval HellaSwag (English), validation split, 5-shot generative MC, accuracy	74.8% ✎	▂··▂▂·▂▆▆█·▄ beats 7/15	medium · avg 63%	▇▇ 5.7%	at its level
Chess Puzzles (Epoch)	0.0% ✎	█▁▁▁▁▂▁▄▂▂▁▁ beats 31/135	very hard · avg 5.9%	▇ 2.0%	sharp · too hard for it · noisy
Vectara hallucination leaderboard (HHEM-2.3, factual consistency rate)	90.7% ✎	▁▂▁▂▅█▂▃▄▃▁▁ beats 51/98	very easy · avg 90%	▇ 0.0%	blunt
GPQA (subset unspecified) lab	75.2% ✎	▂▁▂▂▄▃▆▃▅█▄▃ beats 87/104	very hard · avg 33%	—	not counted yet: no other model in the same run to compare it with

1 more test

test	result	vs the others in its run	difficulty	counts	why
Humanity's Last Exam lab	14.4% ✎	▁▁▁▃▆▆▇██▆▁▃ beats 33/104	very hard · avg 3.8%	—	not counted yet: no other model in the same run to compare it with

$ where to run it

Through an API

api · $0.0000/M · $0 in / $0 out per 1M tokens · 200k context · docs.z.ai ↗ · price sourceStandard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ.

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run GLM-4.7-Flash · needs ~18.1 GB · from huggingface.co/lmstudio-community
chat in the terminal
llama cli -hf lmstudio-community/GLM-4.7-Flash-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf lmstudio-community/GLM-4.7-Flash-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run GLM-4.7-Flash · needs ~18.1 GB · from huggingface.co/lmstudio-community
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/GLM-4.7-Flash-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run GLM-4.7-Flash · needs ~19.22 GB · from ollama.com
chat in the terminal
ollama run glm-4.7-flash:q4_K_M

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run GLM-4.7-Flash · needs ~16.9 GB, Mac only · from huggingface.co/lmstudio-community
chat in the terminal
mlx_lm.chat --model lmstudio-community/GLM-4.7-Flash-MLX-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model lmstudio-community/GLM-4.7-Flash-MLX-4bit

Weights and documentation: huggingface.co/zai-org ↗

$ how the score was worked out · Epoch AI — internal evaluatio… 43%, Roblox LLM Leaderboard (Torpe… 19%, LMArena leaderboard dataset 15% and 3 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Epoch AI — internal evaluations	3	▇▇▇▇▇▇▇▇▇▇▇▇▇ 43%	CC BY 4.0
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode	2	▇▇▇▇▇▇ 19%	MIT
LMArena leaderboard dataset	1	▇▇▇▇ 15%	CC BY 4.0
EuroEval leaderboards (results collection), English generative leaderboard	2	▇▇▇▇ 14%	MIT
EQ-Bench 3 canonical leaderboard results	1	▇▇▇ 9%	MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 1%
weighted average		= 58.8

not counted (7)

zai-org/GLM-4.7-Flash model card	AIME 2025 [correct]	91.6%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	BrowseComp [correct]	42.8%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	GPQA (subset unspecified) [correct]	75.2%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	Humanity's Last Exam [correct]	14.4%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	LiveCodeBench v6 (window unspecified) [correct]	64.0%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	SWE-bench Verified [correct]	59.2%	fewer than three models have this test yet, or this source ran it for this model alone
zai-org/GLM-4.7-Flash model card	τ²-bench aggregate [correct]	79.5%	fewer than three models have this test yet, or this source ran it for this model alone

$ published speed measurements · 14

Reference results for the hardware, engine and settings below. They aren't verified speeds of the downloads above or predictions for your machine.

NVIDIA DGX Spark (GB10)	Q8_0	48.68 tok/s	llama.cpp 11fb327bf (7941), CUDA · llama-bench; tg32; 32 generated tokens; context depth 0; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
NVIDIA DGX Spark (GB10)	Q8_0	44.62 tok/s	llama.cpp 11fb327bf (7941), CUDA · llama-bench; tg32 @ d4096; 32 generated tokens; context depth 4096; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
NVIDIA DGX Spark (GB10)	Q8_0	32.65 tok/s	llama.cpp 11fb327bf (7941), CUDA · llama-bench; tg32 @ d32768; 32 generated tokens; context depth 32768; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
Apple M2 Ultra (GPU core count and RAM not specified)	Q8_0	59.58 tok/s	llama.cpp b828e18c7 (7948), Metal/BLAS · llama-bench; tg32; 32 generated tokens; context depth 0; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
Apple M2 Ultra (GPU core count and RAM not specified)	Q8_0	47.44 tok/s	llama.cpp b828e18c7 (7948), Metal/BLAS · llama-bench; tg32 @ d4096; 32 generated tokens; context depth 4096; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
Apple M2 Ultra (GPU core count and RAM not specified)	Q8_0	21.81 tok/s	llama.cpp b828e18c7 (7948), Metal/BLAS · llama-bench; tg32 @ d32768; 32 generated tokens; context depth 32768; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
NVIDIA DGX Spark (GB10)	Q8_0	42.59 tok/s	llama.cpp 11fb327bf (7941), CUDA · llama-bench; tg32 @ d8192; 32 generated tokens; context depth 8192; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
NVIDIA DGX Spark (GB10)	Q8_0	38.72 tok/s	llama.cpp 11fb327bf (7941), CUDA · llama-bench; tg32 @ d16384; 32 generated tokens; context depth 16384; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
Apple M2 Ultra (GPU core count and RAM not specified)	Q8_0	40.2 tok/s	llama.cpp b828e18c7 (7948), Metal/BLAS · llama-bench; tg32 @ d8192; 32 generated tokens; context depth 8192; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
Apple M2 Ultra (GPU core count and RAM not specified)	Q8_0	31.5 tok/s	llama.cpp b828e18c7 (7948), Metal/BLAS · llama-bench; tg32 @ d16384; 32 generated tokens; context depth 16384; n_ubatch=2048; flash attention=1; one sequence Published token-generation mean; not prompt processing or aggregate batch throughput. Hardware core count and installed RAM are not inferred.	source ↗
NVIDIA GeForce RTX 5090 (32GB; host CPU/RAM not stated)	GGUF Q6_K_XL (Unsloth repository)	151.3 tok/s	llama.cpp (GGUF; version not stated) · Offload column: 0%; prompt-length variant not stated in Table 5 (study prompts ~16/~300/~496 tokens); fixed 256 output tokens, greedy decoding, fixed seed; two untimed warm-up generations; mean of three timed runs; generation throughput measured strictly during the decode phase Model fully resident in GPU/unified memory as published.	source ↗
Apple M4 Pro (14-core CPU / 20-core GPU, 64GB)	MLX 4-bit	53.6 tok/s	MLX v0.30.6 (macOS 26 Tahoe) · Offload column: 0%; prompt-length variant not stated in Table 5 (study prompts ~16/~300/~496 tokens); fixed 256 output tokens, greedy decoding, fixed seed; two untimed warm-up generations; mean of three timed runs; generation throughput measured strictly during the decode phase Model fully resident in GPU/unified memory as published.	source ↗
Apple M3 Ultra (28-core CPU / 60-core GPU, 96GB)	MLX 4-bit	55 tok/s	MLX v0.30.6 (macOS 26 Tahoe) · Offload column: 0%; prompt-length variant not stated in Table 5 (study prompts ~16/~300/~496 tokens); fixed 256 output tokens, greedy decoding, fixed seed; two untimed warm-up generations; mean of three timed runs; generation throughput measured strictly during the decode phase Model fully resident in GPU/unified memory as published.	source ↗
Apple M2 Max (12-core CPU / 38-core GPU, 32GB)	MLX 4-bit	39.7 tok/s	MLX v0.30.6 (macOS 26 Tahoe) · Offload column: 0%; prompt-length variant not stated in Table 5 (study prompts ~16/~300/~496 tokens); fixed 256 output tokens, greedy decoding, fixed seed; two untimed warm-up generations; mean of three timed runs; generation throughput measured strictly during the decode phase Model fully resident in GPU/unified memory as published.	source ↗

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json