Skip to content
Weights Not Walls
Menu

zai-org

GLM-5.3

Reviewed
Size
753B parameterscounted from the safetensors header on the Hub
Licence
unknown, GLM-5.3 LicenseOur reading, check the licence: Commercial use is allowed for almost everyone: only a licensee whose group runs a Model as a Service business with combined revenue over 10 billion US dollars in any consecutive 12 months must first pass a security review whose scope Z.AI decides. Otherwise it is MIT style: keep the copyright and permission notice, with no duty to share changes or pass on other terms. The text is given in English and Chinese.
On the Hub since
25 Aug 2026the date the repo was created
Context length
1M tokensfrom the model's config
Kind
Base model (no base model declared on the Hub)GlmMoeDsaForCausalLM
On Hugging Face
zai-org/GLM-5.31.6M downloads in the last 30 days, 2.5k likes

Scores

23 independent results on 20 eval sets
Chess PuzzlesRe-run by Epoch AI
  • effort max37.0%±4.913 Oct 2026, log

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_max, 79.0% (Anthropic). All 166 results

  • effort max29.3%±7.225 Aug 2026, Epoch AI

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_max, 100.0% (Anthropic). All 83 results

  • effort max68.8%±2.725 Aug 2026, Epoch AI

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_max, 95.1% (Anthropic). All 127 results

GPQA diamondRe-run by Epoch AI
  • effort max90.9%±1.624 Aug 2026, log

Epoch's name: glm-5.3. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results

Mystery Game PuzzlesRe-run by Epoch AI
  • effort max33.0%±4.730 Aug 2026, Epoch AI

Epoch's name: glm-5.3. Best closed model on this set: gpt-6.2-astra_max, 95.0% (OpenAI). All 148 results

OTIS Mock AIME 2024-2025Re-run by Epoch AI
  • effort max91.1%±3.524 Aug 2026, log

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results

SimpleQA VerifiedRe-run by Epoch AI
  • effort max41.0%±1.628 Aug 2026, log

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_max, 94.0% (Anthropic). All 105 results

APEX-AgentsReported

Epoch's name: glm-5.3. Best closed model on this set: gemini-4-carbon_unknown, 84.2% (Google DeepMind). All 66 results

CritPtReported

Epoch's name: glm-5.3. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results

CursorBenchReported

Epoch's name: glm-5.3. Best closed model on this set: gemini-4-carbon_high, 60.4% (Google DeepMind). All 94 results

DeepSWEReported

Epoch's name: glm-5.3. Best closed model on this set: gpt-6-astra_xhigh, 74.1% (OpenAI). All 69 results

DTBenchReported

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results

FrontierCodeReported
  • effort max; Harness: devin-cli; Reasoning effort: max40.1%cognition.com

Epoch's name: glm-5.3. Best closed model on this set: claude-opus-5-5_medium, 54.6% (Anthropic). All 46 results

FrontierSWEReported

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_max, 69.9% (Anthropic). All 25 results

GDP.pdfReported

Epoch's name: glm-5.3. Best closed model on this set: gemini-4-carbon_high, 56.4% (Google DeepMind). All 19 results

LMCAReported

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-1_max, 44.6% (Anthropic). All 22 results

ProofBenchReported

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-1_max, 100.0% (Anthropic). All 85 results

SimpleBenchReported

Epoch's name: glm-5.3. Best closed model on this set: claude-fable-5-2_unknown, 94.9% (Anthropic). All 125 results

WeirdMLReported

Epoch's name: glm-5.3. Best closed model on this set: gpt-6-astra_promax, 93.6% (OpenAI). All 173 results

Claimed by the maker Claimed

The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.

Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.

Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.

Run it yourself

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.69 GB, worked out from the model's config (zai-org/GLM-5.3), at 16-bit precision. Choose a machine above to see a verdict for each format. Method

GGUF for llama.cpp, Ollama, LM Studio and others

unsloth/GLM-5.3-GGUF (converted by unsloth), 332k downloads. Sizes are the real files on the Hub, all parts added up.

FormatSizeMemory neededGet it
UD-IQ1_S201.8 GBfile size, 6 parts203.0 GB201.8 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-IQ1_S
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-IQ1_S/GLM-5.3-UD-IQ1_S-*-of-00006.gguf" --local-dir GLM-5.3-GGUF
UD-IQ1_M212.8 GBfile size, 6 parts214.0 GB212.8 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-IQ1_M
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-IQ1_M/GLM-5.3-UD-IQ1_M-*-of-00006.gguf" --local-dir GLM-5.3-GGUF
UD-IQ2_M222.2 GBfile size, 6 parts223.4 GB222.2 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-IQ2_M
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-IQ2_M/GLM-5.3-UD-IQ2_M-*-of-00006.gguf" --local-dir GLM-5.3-GGUF
UD-Q2_K_XL236.4 GBfile size, 7 parts237.6 GB236.4 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q2_K_XL
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-Q2_K_XL/GLM-5.3-UD-Q2_K_XL-*-of-00007.gguf" --local-dir GLM-5.3-GGUF
UD-IQ3_XXS262.3 GBfile size, 7 parts263.5 GB262.3 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-IQ3_XXS
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-IQ3_XXS/GLM-5.3-UD-IQ3_XXS-*-of-00007.gguf" --local-dir GLM-5.3-GGUF
UD-Q3_K_XL319.4 GBfile size, 9 parts320.6 GB319.4 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q3_K_XL
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-Q3_K_XL/GLM-5.3-UD-Q3_K_XL-*-of-00009.gguf" --local-dir GLM-5.3-GGUF
UD-IQ4_XS340.2 GBfile size, 9 parts341.4 GB340.2 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-IQ4_XS
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-IQ4_XS/GLM-5.3-UD-IQ4_XS-*-of-00009.gguf" --local-dir GLM-5.3-GGUF
UD-Q4_K_XL435.2 GBfile size, 11 parts436.4 GB435.2 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q4_K_XL
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-Q4_K_XL/GLM-5.3-UD-Q4_K_XL-*-of-00011.gguf" --local-dir GLM-5.3-GGUF
UD-Q5_K_XL523.8 GBfile size, 13 parts525.0 GB523.8 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q5_K_XL
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-Q5_K_XL/GLM-5.3-UD-Q5_K_XL-*-of-00013.gguf" --local-dir GLM-5.3-GGUF
UD-Q6_K_XL637.4 GBfile size, 16 parts638.6 GB637.4 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q6_K_XL
Download
hf download unsloth/GLM-5.3-GGUF --include "UD-Q6_K_XL/GLM-5.3-UD-Q6_K_XL-*-of-00016.gguf" --local-dir GLM-5.3-GGUF
Q8_0746.3 GBfile size, 17 parts747.5 GB746.3 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:Q8_0
Download
hf download unsloth/GLM-5.3-GGUF --include "Q8_0/GLM-5.3-Q8_0-*-of-00017.gguf" --local-dir GLM-5.3-GGUF
BF161404.4 GBfile size, 33 parts1405.6 GB1404.4 + 0.69 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/GLM-5.3-GGUF:BF16
Download
hf download unsloth/GLM-5.3-GGUF --include "BF16/GLM-5.3-BF16-*-of-00033.gguf" --local-dir GLM-5.3-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 6bitAinsleyMarkov/GLM-5.3-MLX-6bit-MTP633.3 GBfile size634.4 GB633.3 + 0.69 + 0.50Download page
MLXorcarouter/GLM-5.3-MLX2122.4 GBfile size2123.6 GB2122.4 + 0.69 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
F8_E4M3original weights703.6 GBfile size704.7 GB703.6 + 0.69 + 0.50
hf download zai-org/GLM-5.3 --local-dir GLM-5.3
Other conversions (20; sizes not read)

The Hub tags 100 or more quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.

Run it by API

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

34 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.

ProviderInOutContextPrecisionTokens/sFirst tokenUptime 24hToolsJSON
Featherless$0.42$1.32256knot given751.5 s99.3%yesno
Wafer$0.07$1.401Mfp4780.4 s100.0%yesyes
Decart$0.56$1.811Mfp41020.6 s100.0%yesyes
SiliconFlow$0.70$1.901Mfp8562.4 s99.9%yesno
Novita$0.63$1.981Mfp8502.1 s100.0%yesyes
Cloudflare$0.70$2.201Mnot given812.5 s99.9%yesyes
DeepInfra$0.56$2.501Mfp4691.0 s99.9%yesyes
Tenstorrent$0.60$2.501020kfp81130.7 s100.0%yesno
DigitalOcean$0.91$2.861Mnot given920.6 s100.0%yesyes
CoreWeave$0.75$3.001Mnvfp41030.5 s100.0%yesyes
Sail Research$0.75$3.001Mfp8620.9 s100.0%yesyes
Fireworks$0.98$3.081Mnot given760.6 s99.6%yesyes
Morph$0.67$3.131Mfp8691.0 s100.0%yesyes
Inceptron$1.05$3.401Mfp4880.7 s99.8%yesno
Cohere$0.50$3.501Mnvfp4850.7 s100.0%yesyes
InferenceNet$0.084$3.601Mfp4940.4 s100.0%yesyes
Relace$0.07$4.381Mfp4701.2 s100.0%yesno
AkashML$0.19$4.401Mfp41191.9 s99.9%yesyes
BaseTen$1.40$4.401Mfp4741.3 s100.0%yesyes
Crusoe$1.40$4.401Mfp4830.6 s100.0%yesyes
Doubleword$1.40$4.401Mnvfp4870.7 s100.0%yesyes
Makora$1.50$4.401Mfp41720.5 s99.1%yesyes
Mistral$1.40$4.401Mnvfp4861.6 s99.8%yesyes
Modal$1.40$4.401Mnot given840.3 s99.9%yesyes
Nebius$1.40$4.401Mfp41361.3 s100.0%yesyes
nscale$1.40$4.401Mnot given1160.6 s?yesyes
Parasail$1.40$4.401Mfp41320.3 s100.0%yesyes
SambaNova$1.40$4.401Mnot given872.2 s100.0%yesyes
Together$1.40$4.401024knot given860.5 s100.0%yesyes
Venice$1.40$4.40977knot given1070.7 s100.0%yesyes
Z.AIthe maker$1.40$4.401Mfp8562.4 s100.0%yesyes
Amazon Bedrock$1.68$5.281Mnot given910.9 s99.6%yesyes
ModelRun$2.10$6.601Mfp41090.3 s100.0%yesyes
zai-orgthe makerserves zai-org/GLM-5.3-BF16???not given352.5 s?yesyes

Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.

curl https://router.huggingface.co/v1/chat/completions \ -H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \ -d '{"model": "zai-org/GLM-5.3:featherless-ai", "messages": [{"role": "user", "content": "Hello"}]}'

Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.

Fine-tunes of this model

from our snapshot of the Hub

Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.

Planned for this page Not built yet

Try it in private chat
A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
Fine-tune it
Fine-tuning a few base models from this page and publishing the result here.
Creators and earnings
Pages for people who publish fine-tunes and adapters of this model, and what they earn.
Fund and bounties
Bounties for new open models, paid from a public fund. How the fund will work.
Contribute compute
How to give compute to open model work, with Akash.

Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.