Model Follower.

Up to date

Rebuilt
2026-10-26 07:36 UTC
Release radar
116 on watch
Hub snapshot
2026-10-26
Re-checked
daily · every number sourced and dated

Models › GLM-4.7 family

GLM-4.7

Hugging Face ↗ Follow changes

zai-org MIT out 10 months ago

358B parameters (8 of 160 experts per token), text in, text out, 198K tokens of context. 7 providers serve it from $0.32 per million input tokens.

Verified 2026-10-26 07:27 UTC Rank #82 in LLMs · score 103 method v1 Sources 4 Downloads 30d 77.9k · #341 of 619 tracked

Changed this week: Cheapest input price fell 20%: $0.400 → $0.320 per 1M (OpenRouter).2026-10-24

Can I run it?215GB at INT4Estimateneeds a multi-GPU nodemethodCommunity: GLM-4.7: Specifications and GPU VRAM Requirements →
What does it cost?$0.32/1M incheapest of 5DeepInfrafp42026-10-26
Can I use it?MITPermissivelicence file2026-10-26
Do people use it?77.9kdownloads, 30 days#341 of 619 we track · below the median+4 likes, downloads −6% this week2,072 likesHugging Face2026-10-26

Worth reading

Picked by rule from 20 tracked pieces: the lab's own post, the most shared, a field report, the latest.

GLM-4.7 achieves 42.8% on the HLE benchmark with tools, a specific improvement over its 30.4% score without them.

From GLM-4.7: Advancing the Coding Capability · z.ai · 2025-12-22
The lab's own post

GLM-4.7: Advancing the Coding Capability

Z.ai released GLM-4.7, which supports thinking before acting to improve performance in complex tasks within mainstream agent frameworks.

  • GLM-4.7 achieves 42.8% on the HLE benchmark with tools, a specific improvement over its 30.4% score without them.
  • The model produces cleaner webpages and generates slides with more accurate layout and sizing for UI tasks.
z.ai 2025-12-22 437 HN pts · 235 comments 2.9k words
Most discussed · 16 HN pts

Open-Weight Models Are Getting Serious: GLM 4.7 vs. MiniMax M2.1

GLM 4.7 and MiniMax M2.1 successfully built a CLI task runner with 20 features in a multi-phase test.

  • MiniMax M2.1 used batch-based execution where tasks are grouped by dependency level.
  • GLM 4.7 implemented argument parsing manually to avoid external dependencies.
blog.kilo.ai 2026-01-09 16 HN pts 1.5k words

20 best of 20 tracked

The 5 picks above, plus these 15. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. ArticleGLM-4.7docs.z.ai · 2025-12-22 · 25 HN pts · official
  2. ArticleUnsloth: GLM-4.7-Flashunsloth.ai · 2026-01-20 · 4 HN pts
  3. ArticleGLM-4.7: Frontier intelligence at record speed – now available on Cerebrascerebras.ai · 2026-01-08 · 9 HN pts
  4. CodeGLM Proxy: transforms Anthropic Messages API requests to Z.ai GLM-4.7 API formatgithub.com · 2025-12-27 · 2 HN pts
  5. ArticleGLM-4.7 · LLM Model Review | CrucibleMarkcruciblemark.com · undated
  6. ArticleA Technical Analysis of GLM-4.7 - Mediummedium.com · undated
  7. ArticleGLM-4.7: Specifications and GPU VRAM Requirementsapxml.com · undated
  8. ArticleGLM-4.7 Specs & Benchmarks | TheModelverse | The Modelversethemodelverse.in · undated
  9. ArticleGLM-4.7: Z.ai's Frontier Agentic Reasoning Model - Elliot Arledgeelliotarledge.com · undated
  10. ArticleGLM-4.7-Flash: 30B MoE model achieves 59.2% on SWE-bench, runs on 24GB GPUscurateclick.com · 2026-01-20
  11. VideoGLM4.7-Flash vs. other Local LLM. Tested also with OpenCode [video]youtube.com · 2026-01-25
  12. ArticleGLM-4.7-Flash 30B-A3B MoExcancel.com · 2026-01-19
  13. ArticleGLM-4.7: The High-Performance Coding Alternative | Differdiffer.blog · 2026-01-31
  14. ArticleGLM4.7-Flash the new Local LLM king at 30B A3B and OpenCode?grigio.org · 2026-01-25
  15. ArticleGLM 4.7: A Complete Deep Dive into Zhipu AI's Flagship Coding Model | AiCybr Blogaicybr.com · 2026-01-03

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

Prices for the same weights vary 2×

5 providers with a live price, from $0.32 to $0.70 per million input tokens; the median is $0.54. One dot per provider, cheapest in teal.

$0.010$0.10$1.00$10.00DeepInfra: $0.32 in / $1.40 out per 1M · observed 2026-10-26 · https://huggingface.co/api/models/zai-org/GLM-4.7?expand[]=inferenceProviderMappingVenice: $0.40 in / $1.93 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7Novita: $0.54 in / $1.98 out per 1M · observed 2026-10-26 · https://huggingface.co/api/models/zai-org/GLM-4.7?expand[]=inferenceProviderMappingZ.AI: $0.60 in / $2.20 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7Mancer 2: $0.70 in / $2.50 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7cheapest $0.32median $0.54
$0.010$0.10$1.00$10.00DeepInfra: $0.32 in / $1.40 out per 1M · observed 2026-10-26 · https://huggingface.co/api/models/zai-org/GLM-4.7?expand[]=inferenceProviderMappingVenice: $0.40 in / $1.93 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7Novita: $0.54 in / $1.98 out per 1M · observed 2026-10-26 · https://huggingface.co/api/models/zai-org/GLM-4.7?expand[]=inferenceProviderMappingZ.AI: $0.60 in / $2.20 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7Mancer 2: $0.70 in / $2.50 out per 1M · observed 2026-10-26 · https://openrouter.ai/z-ai/glm-4.7cheapest $0.32 · median $0.54 (white tick)
#Cheapest 5In /1MOut /1MContextQuantSeen
1 DeepInfra $0.32 $1.40 198K fp4 Oct 26
2 Venice $0.40 $1.93 193K fp4 Oct 26
3 Novita $0.54 $1.98 200K fp8 Oct 26
4 Z.AI $0.60 $2.20 198K fp4 Oct 26
5 Mancer 2 $0.70 $2.50 128K fp4 Oct 26

All 7 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Widely served (7 API providers); 43 quantized builds for local use.

  • As a service7DeepInfra, Featherless, Google, Mancer 2 +3
  • On own hardware43quantized builds
  • Built on3021 fine-tunes · 9 adapters · 0 merges
  • In apps100+Hugging Face Spaces

Top Spaces running it: enzostvs/deepsite ♥ 16,604 · akhaliq/anycoder ♥ 3,321 · hadadxyz/ai ♥ 819 · pliny-the-prompter/obliteratus ♥ 621

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Needs an 8×80 GB node even at INT4

About 215 GB at INT4, 430 GB at FP8, 860 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 215 GB of weights (estimate; KV cache and activations extra)INT4215 GBFP8: about 430 GB of weights (estimate; KV cache and activations extra)FP8430 GBBF16: about 860 GB of weights (estimate; KV cache and activations extra)BF16860 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 215 GB of weights (estimate; KV cache and activations extra)INT4215 GBFP8: about 430 GB of weights (estimate; KV cache and activations extra)FP8430 GBBF16: about 860 GB of weights (estimate; KV cache and activations extra)BF16860 GB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

7 dated articles and 14 repo commits so far

Teal is the weights (2025-12-22), amber triangles are 7 dated articles (each links to it), grey ticks are repo commits.

JanFebMarAprMayJunJulAugSepOct2026-01-29: 3 repo commits2026-01-08: 1 repo commit2025-12-23: 3 repo commits2025-12-22: 7 repo commitsGLM-4.7: Advancing the Coding Capability · 2025-12-22GLM-4.7 · 2025-12-22Claude Code Uses GLM 4.7 · 2026-04-27Unsloth: GLM-4.7-Flash · 2026-01-20Open-Weight Models Are Getting Serious: GLM 4.7 vs. MiniMax M2.1 · 2026-01-09GLM-4.7: Frontier intelligence at record speed – now available on Cerebras · 2026-01-08GLM Proxy: transforms Anthropic Messages API requests to Z.ai GLM-4.7 API format · 2025-12-27Weights 2025-12-22 · 2025-12-22Weights 2025-12-22now · 7 providers
JanFebMarAprMayJunJulAugSepOct2026-01-29: 3 repo commits2026-01-08: 1 repo commit2025-12-23: 3 repo commits2025-12-22: 7 repo commitsGLM-4.7: Advancing the Coding Capability · 2025-12-22GLM-4.7 · 2025-12-22Claude Code Uses GLM 4.7 · 2026-04-27Unsloth: GLM-4.7-Flash · 2026-01-20Open-Weight Models Are Getting Serious: GLM 4.7 vs. MiniMax M2.1 · 2026-01-09GLM-4.7: Frontier intelligence at record speed – now available on Cerebras · 2026-01-08GLM Proxy: transforms Anthropic Messages API requests to Z.ai GLM-4.7 API format · 2025-12-27Weights 2025-12-22 · 2025-12-22Weights 2025-12-22now · 7 providers

What people built on it

73 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

73 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

43 quantized (Hugging Face model tree)21 fine-tunes (Hugging Face model tree)9 adapters (Hugging Face model tree)43 quantized21 fine-tunes9 adapters
43 quantized (Hugging Face model tree)21 fine-tunes (Hugging Face model tree)9 adapters (Hugging Face model tree)43 quantized21 fine-tunes9 adapters
GGUF unsloth/GLM-4.7-GGUF for llama.cpp, Ollama, LM Studio 386.4k downloads MLX mlx-community/GLM-4.7-4bit for Apple silicon 95.8k downloads AWQ QuantTrio/GLM-4.7-AWQ for vLLM, SGLang 145.8k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

17 rows

Benchmarks 17 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (17)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
BenchmarkValueSetting / notesReported byDate
author-reported AIME 2025 · Reason through hard problems95.7—model author2026-10-07
author-reported BrowseComp52—model author2026-10-07
author-reported GPQA Diamond · Reason through hard problems85.7—model author2026-10-07
author-reported HLE (w/ Tools)42.8—model author2026-10-07
author-reported HMMT · Reason through hard problems97.1—model author2026-10-07
author-reported HMMT · Reason through hard problems93.5—model author2026-10-07
author-reported Humanity's Last Exam · Reason through hard problems24.8—model author2026-10-07
author-reported IMO-AnswerBench · Reason through hard problems82—model author2026-10-07
author-reported LiveCodeBench · Write code with agents84.9—model author2026-10-07
author-reported MMLU-Pro · Reason through hard problems84.3—model author2026-10-07
author-reported SWE-bench Multilingual · Write code with agents66.7—model author2026-10-07
author-reported SWE-bench Verified · Write code with agents73.8—model author2026-10-07
author-reported Terminal-Bench · Write code with agents33.3—model author2026-10-07
author-reported Terminal-Bench · Write code with agents41—model author2026-10-07
author-reported diamond (Idavidrein/gpqa)85.7—Model Card2026-01-21
author-reported hle (cais/hle)24.8—Model Card2026-01-14
author-reported TIGER-Lab/MMLU-Pro84.3—Model Card2026-01-15

Use-case evidence

5 use cases

What it is good at 5 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

24 fields

Listed in LLMs · MoE

Facts

Pipeline tagtext-generation
Librarytransformers
ArchitecturesGlm4MoeForCausalLM
Model typeglm4_moe
Dtypebfloat16
Params by dtypeBF16: 358,337,776,896 · F32: 14,400
Size bandXXL
MoEyes
Experts160
Experts per token8
Context length202,752 (config.json max_position_embeddings)
Languagesen, zh
PapersarXiv:2508.06471
LicenseMIT (mit)
Hub snapshot2026-10-26

Servability

vLLM · SGLang

Servability

vLLM: supported · Glm4MoeForCausalLM · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · Glm4MoeForCausalLM · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f

vLLM registry (4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60) · SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for GLM-4.7: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)860 GB · Cluster (≥640 GB)
FP8 (1 byte/param)430 GB · Multi-GPU node (<640 GB)
INT4 (0.5 bytes/param)215 GB · Multi-GPU node (<640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original model Family: zai-org/GLM-4.7 73 derivatives on the Hub

quantized: 43finetune: 21adapter: 9

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
unsloth/GLM-4.7-GGUF quantized GGUF — — mit unsloth 225 386.4k
QuantTrio/GLM-4.7-AWQ quantized AWQ 358B XXL mit QuantTrio 28 145.8k
mlx-community/GLM-4.7-4bit quantized -4bit 353B XXL mit mlx-community 5 95.8k
mlx-community/GLM-4.7-6bit quantized — 353B XXL mit mlx-community 1 81.9k
Tengyunw/GLM-4.7-NVFP4 quantized NVFP4 177B XXL mit Tengyunw 6 44.9k
bartowski/zai-org_GLM-4.7-GGUF quantized GGUF — — mit bartowski 6 35.8k
Salyut1/GLM-4.7-NVFP4 quantized NVFP4 177B XXL mit Salyut1 16 32.5k
nvidia/GLM-4.7-NVFP4 quantized NVFP4 183B XXL mit nvidia 19 29.1k
Ex0bit/GLM-4.7-PRISM quantized GGUF 353B XXL mit Ex0bit 30 11.4k
AaryanK/GLM-4.7-GGUF quantized GGUF — — mit AaryanK 14 10.1k
lmstudio-community/GLM-4.7-GGUF quantized GGUF — — mit lmstudio-community 1 9.1k
ubergarm/GLM-4.7-GGUF quantized GGUF — — mit ubergarm 26 6.5k
soundsgoodai/GLM-4.7-NVFP4-KV-cache-FP8 quantized NVFP4 177B XXL apache-2.0 soundsgoodai 1 7.4k
cerebras/GLM-4.7-REAP-218B-A32B finetune — 218B XXL mit cerebras 33 4.5k
DevQuasar/zai-org.GLM-4.7-GGUF quantized GGUF — — — DevQuasar 0 5.6k
QuantPasture/GLM-4.7-GGUF quantized GGUF — — — QuantPasture 14 4.7k
cyankiwi/GLM-4.7-AWQ-4bit quantized AWQ 59.1B XL mit cyankiwi 6 4.7k
mlx-community/GLM-4.7-8bit quantized -8bit 353B XXL mit mlx-community 3 4.5k
mradermacher/GLM-4.7-i1-GGUF quantized GGUF — — mit mradermacher 4 3.9k
Intel/GLM-4.7-int4-mixed-AutoRound quantized int4 1.63B S — Intel 26 2.1k
amd/GLM-4.7-MXFP4 quantized MXFP4 184B XXL mit amd 3 2.7k
jarradh/GLM-4.7-heretic finetune — 353B XXL mit jarradh 8 2k
unsloth/GLM-4.7-FP8 quantized -FP8 358B XXL mit unsloth 3 2.1k
soundsgoodai/GLM-4.7-NVFP4-KV-cache-BF16 quantized NVFP4 177B XXL apache-2.0 soundsgoodai 0 2.3k
unsloth/GLM-4.7 finetune — 358B XXL mit unsloth 10 1.6k
QuantTrio/GLM-4.7-GPTQ-Int4-Int8Mix quantized GPTQ 390B XXL mit QuantTrio 4 1.7k
mratsim/GLM-4.7-EXL3 quantized EXL3 — — mit mratsim 21 497
cerebras/GLM-4.7-REAP-268B-A32B finetune — 269B XXL mit cerebras 19 575
mradermacher/GLM-4.7-GGUF quantized GGUF — — mit mradermacher 2 1.4k
noctrex/GLM-4.7-MXFP4_MOE-GGUF quantized GGUF — — — noctrex 5 1.1k
mrtoots/GLM-4.7-mlx-6Bit quantized mlx 353B XXL mit mrtoots 0 995
gghfez/GLM-4.7-Q8-GGUF quantized GGUF — — — gghfez 1 827
jkim96/GLM-4.7-DASHQ-INT2-g32 quantized — 1.63B S mit jkim96 0 824
Jon-Nielsen/GLM-4.7-REAP-30-W4A16 quantized -W4A16 1.60B S apache-2.0 Jon-Nielsen 3 553
lovedheart/GLM-4.7-GGUF-IQ1_M quantized GGUF — — — lovedheart 5 452
jesusoctavioas/GLM-4.7-mlx-4Bit quantized mlx 353B XXL mit jesusoctavioas 0 698
jesusoctavioas/GLM-4.7-mlx-2Bit quantized mlx 353B XXL mit jesusoctavioas 0 602
FramedStone/GLM-4.7-Q4 quantized — 353B XXL mit FramedStone 0 589
gghfez/GLM-4.7-Derestricted-v3-Q3KL-GGUF quantized GGUF — — mit gghfez 5 325
ConicCat/GLM-4.7-Architect-355B-A32B finetune — — — mit ConicCat 7 155

Repository changelog

14 commits
changed in the last 7 daysPrice moves: 2026-10-24all changes

Repository changelog 14 commits

DateCommitSHA
2026-01-29 Add GPQA Eval Result (#37) 602d01e
2026-01-29 Add MMLU-Pro evaluation result (#34) e2ee390
2026-01-29 Add HLE evaluation result (#31) 9258c1a
2026-01-08 Update README.md: remove `extra_body` in disabling thinking (#27) 475d85c
2025-12-23 add ed30813
2025-12-23 update 49342a6
2025-12-23 readme 26f32ad
2025-12-22 fix readme 9fcd12a
2025-12-22 new readme 2942382
2025-12-22 Update README.md 13c375b
2025-12-22 Update README.md 4f271a7
2025-12-22 Add files using upload-large-folder tool 646847c
2025-12-22 Add files using upload-large-folder tool d2aa2a8
2025-12-22 initial commit d336a2b

All providers

7 providers · 7 endpoints

Where to run 7 providers · 7 endpoints

Hosted endpoints observed for GLM-4.7, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.32/M.

ProviderModel refContextQuant$/M in$/M outtok/sTTFTToolsStatusObserved
DeepInfra
hf, deepinfra, openrouter
 zai-org/GLM-4.7 198K fp4 $0.32 $1.40 25.8 361 ms yes live 2026-10-26
Venice
openrouter
 z-ai/glm-4.7 193K fp4 $0.40 $1.93 — — yes live 2026-10-26
Novita
hf, openrouter
 zai-org/glm-4.7 200K fp8 $0.54 $1.98 27 3735 ms yes live 2026-10-26
Google
openrouter
 z-ai/glm-4.7 195K unknown $0.60 $2.20 — — yes status -5 2026-10-26
Z.AI
openrouter
 z-ai/glm-4.7 198K fp4 $0.60 $2.20 — — yes live 2026-10-26
Mancer 2
openrouter
 z-ai/glm-4.7 128K fp4 $0.70 $2.50 — — — live 2026-10-26
Featherless
hf
 zai-org/GLM-4.7 — — — — — — — live 2026-10-26

Articles and coverage

20 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official4

GLM-4.7: Advancing the Coding Capability
z.ai2025-12-22officialauthorityHN front pageHN 437 pts · 235 comments2,892 words
Z.ai released GLM-4.7, which supports thinking before acting to improve performance in complex tasks within mainstream agent frameworks. The model achieves 73.8% on SWE-bench, 66.7% on SWE-bench Multilingual, and 41% on Terminal Bench 2.0. It delivers a 42.8% score on the HLE benchmark and shows gains across 17 benchmarks compared to GLM-4.6.written, reviewed, cited
  • GLM-4.7 achieves 42.8% on the HLE benchmark with tools, a specific improvement over its 30.4% score without them.
  • The model produces cleaner webpages and generates slides with more accurate layout and sizing for UI tasks.
  • GLM-4.7 achieves significantly better performance on web browsing via BrowseComp compared to other models like Gemini 3.0 Pro.
GLM-4.7
docs.z.ai2025-12-22officialauthorityHN 25 pts · 6 comments949 words
Z.AI released GLM-4.7, a model series featuring enhanced programming capabilities and stable multi-step reasoning. It supports 200K context length and 128K maximum output tokens, offering multiple thinking modes and real-time streaming responses for complex agent tasks. The model scores 67 points on the BrowseComp web task evaluation and achieves an open-source SOTA of 84.7 points on the τ²-Bench interactive tool invocation benchmark. Additionally, it reaches 42.8% on the HLE benchmark, representing a 41% increase over GLM-4.6.written, reviewed, cited
  • GLM-4.7 implements a 'think before acting' mechanism within programming frameworks like Claude Code, Kilo Code, TRAE, Cline, and Roo Code.
  • The model achieves 73.8% on SWE-bench Verified, marking a 5.8% improvement over GLM-4.6.
Unsloth: GLM-4.7-Flash
unsloth.ai2026-01-20authorityHN 4 pts · 0 comments1,214 words
GLM-4.7-Flash is a 30B MoE reasoning model from Z.ai that uses ~3.6B parameters and supports a 202,752 token context window. Running the 4-bit quantized version requires around 18GB RAM, while Unsloth's FP8 Dynamic quant reduces KV cache memory usage by 50% on 4 GPUs. However, using the MTP module for speculative decoding drops generation throughput from 13,000 tokens/s to 1,300 tokens/s on a single B200 GPU.written, reviewed, cited
  • The 30B model does not fit on a free Colab GPU for fine-tuning.
  • Fine-tuning the router layer is disabled by default to preserve reasoning capabilities.
  • Using llama.cpp requires setting --min-p 0.01 because the default is 0.05.
GLM-4.7: Frontier intelligence at record speed – now available on Cerebras
cerebras.ai2026-01-08HN 9 pts · 1 comments748 words
Z.ai released GLM-4.7, an open-weight model available on Cerebras Inference Cloud that achieves approximately 1,000 tokens per second and up to 1,700 TPS for specific use cases. The model delivers roughly 10x higher price-performance than Claude Sonnet 4.5 while maintaining comparable intelligence to DeepSeek-V3.2 on developer benchmarks. Teams can access the model via Cerebras Cloud starting at a $10 pay-as-you-go developer tier.written, reviewed, cited
  • GLM-4.7 introduces preserved thinking, allowing reasoning context to persist across turns.
  • The model interleaves thinking before each action, tool call, or response rather than treating reasoning as a single upfront step.
  • Migration requires updating the model name while keeping the same API surface as GLM-4.6.

Analysis5

GLM-4.7 Models - Intelligence, Performance & Price Comparison | Artificial Analysis
artificialanalysis.aiauthority464 words
Z AI released GLM-4.7, a model family with 357B total parameters and an MIT license. The reasoning variant scores 22 on the Intelligence Index, outputs at 40 tokens per second, and costs $0.7 per 1M tokens. The non-reasoning version achieves a 1.57s latency to the first token and supports a 200k context window with 32B active parameters during inference.written, reviewed, cited
Open-Weight Models Are Getting Serious: GLM 4.7 vs. MiniMax M2.1
blog.kilo.ai2026-01-09HN 16 pts · 0 comments1,537 words
GLM 4.7 and MiniMax M2.1 successfully built a CLI task runner with 20 features in a multi-phase test. GLM 4.7 completed the task in 10 minutes with a 741-line plan and full README, whereas MiniMax M2.1 finished in 14 minutes with a 284-line plan and no documentation. MiniMax M2.1 cost $0.15 for the task and is currently free to use in Kilo Code for a limited time.written, reviewed, cited
  • MiniMax M2.1 used batch-based execution where tasks are grouped by dependency level.
  • GLM 4.7 implemented argument parsing manually to avoid external dependencies.
  • MiniMax M2.1 uses SHA256 for hashing, which is technically more collision-resistant than GLM 4.7's MD5.
GLM Proxy: transforms Anthropic Messages API requests to Z.ai GLM-4.7 API format
github.com2025-12-27authorityHN 2 pts · 1 comments2,986 words
GLM-4.7: Pricing, Benchmarks, and Full Model Analysis
llm-stats.com1,667 words
GLM-4.7 · LLM Model Review | CrucibleMark
cruciblemark.com2,246 words

Community1

Claude Code Uses GLM 4.7
old.reddit.com2026-04-27HN 2 pts · 4 comments
All coverage tracked for this model (20) — append-only, nothing is ever removed
TitleSourcePublishedFirst seenSignals
GLM-4.7 Models - Intelligence, Performance & Price Comparison | Artificial Analysisartificialanalysis.ai—2026-10-07exa
GLM-4.7: Pricing, Benchmarks, and Full Model Analysisllm-stats.com—2026-10-07exa
GLM-4.7 · LLM Model Review | CrucibleMarkcruciblemark.com—2026-10-07exa
A Technical Analysis of GLM-4.7 - Mediummedium.com—2026-10-07exa
GLM-4.7: Specifications and GPU VRAM Requirementsapxml.com—2026-10-07exa
GLM-4.7 Specs & Benchmarks | TheModelverse | The Modelversethemodelverse.in—2026-10-07exa
GLM-4.7: Z.ai's Frontier Agentic Reasoning Model - Elliot Arledgeelliotarledge.com—2026-10-07exa
Claude Code Uses GLM 4.7old.reddit.com2026-04-272026-10-07hn
GLM-4.7: The High-Performance Coding Alternative | Differdiffer.blog2026-01-312026-10-07exa
GLM4.7-Flash vs. other Local LLM. Tested also with OpenCode [video]youtube.com2026-01-252026-10-07hn
GLM4.7-Flash the new Local LLM king at 30B A3B and OpenCode?grigio.org2026-01-252026-10-07hn
Unsloth: GLM-4.7-Flashunsloth.ai2026-01-202026-10-07hn
GLM-4.7-Flash: 30B MoE model achieves 59.2% on SWE-bench, runs on 24GB GPUscurateclick.com2026-01-202026-10-07hn
GLM-4.7-Flash 30B-A3B MoExcancel.com2026-01-192026-10-07hn
Open-Weight Models Are Getting Serious: GLM 4.7 vs. MiniMax M2.1blog.kilo.ai2026-01-092026-10-07hn
GLM-4.7: Frontier intelligence at record speed – now available on Cerebrascerebras.ai2026-01-082026-10-07hn
GLM 4.7: A Complete Deep Dive into Zhipu AI's Flagship Coding Model | AiCybr Blogaicybr.com2026-01-032026-10-07exa
GLM Proxy: transforms Anthropic Messages API requests to Z.ai GLM-4.7 API formatgithub.com2025-12-272026-10-07hn
GLM-4.7: Advancing the Coding Capabilityz.ai official2025-12-222026-10-07hn
GLM-4.7docs.z.ai official2025-12-222026-10-07hn, exa

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Similar models

3 models

Similar models

  • GLM-4.6 · 357B
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
  • DeepSeek-V4-Flash-0731 · 304B
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
  • GLM-4.5 · 358B
    same size band (XXL), also MoE, same task (text-generation), fewer providers, permissive license

Sources

4 sources

Sources

Hub snapshot: 2026-10-26 · HF API: 2026-10-26 07:15 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-26 07:27 UTC

Follow GLM-4.7

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.