whatmodel.sh / GLM-5.3-Flash

GLM-5.3-Flash

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

GLM-5.3-Flash is an open-weights language model from Z.ai, released August 2026. Its published size is 320B parameters (18B active per token) and a 1M-token context window. Licence: mit.

74.0±0.7#49 of 719 LLMsfrom 21 benchmark results across 8 sources

$ how it compares

The LLMs ranked either side of GLM-5.3-Flash, out of 828 with a score, then more from Z.ai.

#modelscorewhereRAMreleased
50DeepSeek V4 Flash (0731) DeepSeek74±1api $0.11/M +local~92GB2026-07
51Qwen3.7 Max Alibaba74±1api $3.75/M2026-05
52GLM-5.3-Flash Z.ai74±1api $0.24/M +local~103GB2026-08
53Gemini 3.5 Flash Google74±1api $3.38/M2026-05
54GPT-5.2 OpenAI74±1api $4.81/M2025-12
more from Z.ai
33GLM-5.3 Z.ai76±1api $2.15/M +local~239GB2026-08
57GLM-5.2 Z.ai74±1api $2.15/M +local~121GB2026-06
81GLM-5.1 Z.ai71±1api $2.15/M +local~119GB2026-04

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #24 of 300 · 22% of its score · 7 tests

Beats GPT-5.5, GPT-5.6 Terra, GPT-5.4 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Arena Agent (October 2026 aggregate IPS)-0.0044 IPS ✎▁▁▁▄▂▃▇█▂▂▁▃ beats 18/49avg -0.312 IPS▇▇▇▇ 9.5%sharp
Terminal-Bench 4.035.8% ✎█▅▅▅▅·▅▃▃▅▅▅ beats 11/20very hard · avg 0.4%▇▇ 4.2%sharp
Terminal-Bench 2.1 lab84.3% ✎▁·▁·▁▁▄▃▄█▃▇ beats 45/56hard · avg 23%▇ 2.6%small run
Video2World V2WScore (coding agents reconstruct interactive simulated worlds from videos, OpenCode scaffold)25.7% ✎█··▄█·▄····▄ beats 0/6very hard · avg 6.6%▇ 2.1%at its level
Humanity's Last Exam with tools lab55.3% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 39/54hard · avg 16%▇ 1.3%at its level · small run
Agents' Last Exam (metric/harness unspecified) lab26.3% ✎▂·▂·█▂···▂▄▄ beats 3/11very hard · avg 6.8%▇ 1.1%at its level · small run
DeepSWE v1.1 (resolved) lab63.4% ✎▂····▁▁▁▆██▆ beats 10/28hard · avg 19%▇ 1.0%at its level · small run
coding▁▁▁▃▆██▇▆▄▃▁ #24 of 458 · 13% of its score · 3 tests

Beats GPT-5.5, GPT-5.6 Terra, Claude Opus 4.8 here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Arena WebDev (Code Arena)1615.6 ✎▁▂▄▃▅█▄▇▆▄▁▁ beats 97/119avg 981▇▇▇ 8.4%sharp
Luau Leetcode fully solved (Torpedo Software)68.8% ✎▁▁▁▁·▁▂▃▅▃█▇ beats 22/49hard · avg 18%▇ 3.4%at its level
DeepSWE v1.1 (resolved) lab63.4% ✎▂····▁▁▁▆██▆ beats 10/28hard · avg 19%▇ 1.0%at its level · small run
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #28 of 393 * · 4.7% of its score · 1 test

Beats Claude Sonnet 5.5, GPT-6 Sol, GPT-5.6 Terra here, though they rank higher overall. Trails GLM-5.2, Gemini 3.6 Flash, Gemini 3.5 Flash here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena (text)1473.0 ✎▁▁▁▂▃▃▃▆▆▇█▃ beats 346/379avg 1268▇▇ 4.7%blunt
math▁▂▃▅▆▅▄█▄▁▁▁ #48 of 398 · 41% of its score · 3 tests

Beats Grok 4.7 here, though they rank higher overall. Trails A.X K2 (688B-A33B), DeepSeek V4 Flash (Preview, Apr 2026), GPT-5 here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
FrontierMath tiers 1–3 v255.8% ✎▂··▁▁▃▄▄█▅▃▃ beats 41/82very hard · avg 1.2%▇▇▇▇▇▇▇▇▇ 23%sharp · at its level
FrontierMath tier 4 v217.1% ✎▁▂▁▂█▅▂▂▁▁▁▁ beats 20/63very hard · avg 0.1%▇▇▇▇▇▇▇ 17%sharp · at its level
OTIS mock AIME 2024–202593.9% ✎▃▃▄▃▃▃▇▇█▃▂▅ beats 157/193very hard · avg 8.5%▇ 1.1%sharp · maxed out · noisy
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #23 of 147 * · 3.3% of its score · 2 tests

Beats Claude Sonnet 5.5, MiMo-V2.6-Pro, GPT-5.6 Terra here, though they rank higher overall. Trails Gemini 3.6 Flash, Gemini 3.5 Flash here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena vision (style controlled)1278.4 ✎▁▁▂▂▂▂▄▅▅▆██ beats 122/145avg 1068▇ 3.3%blunt
PureDocBench document parsing (Avg3 over Clean/Digital/Real)79.7% ✎▂·▁·▂▂▅█·▁▂▃ beats 22/22medium · avg 62%▇ 0.1%blunt · at its level
web search▂▁▂▆▄█▃▃▁▁·▁ #36 of 80 * · 1.3% of its score · 1 test

Beats Gemini 3.1 Pro (Preview), Qwen3.7 Max here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Humanity's Last Exam with tools lab55.3% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 39/54hard · avg 16%▇ 1.3%at its level · small run
other tests6 not in a task · 17% of its score
testresultvs the others in its rundifficultycountswhy
Mystery Game Puzzles (Epoch)8.0% ✎█··▂▃▇▄█▂▃▁▁ beats 15/74very hard · avg 9.4%▇▇ 5.4%sharp · noisy
GPQA Diamond90.2% ✎▁▁▁▁▂▄▆▄▅▇█▅ beats 174/208hard · avg 40%▇▇ 4.0%sharp
Chess Puzzles (Epoch)14.0% ✎█▁▁▁▁▃▁▄▂▂▁▁ beats 68/135very hard · avg 5.9%▇ 3.7%sharp · noisy
Icelandic LLM Leaderboard average (Miðeind, Sept 2026 suite)27.9% ✎▁▁▃▂▂▄▁▃▄▄▄█ beats 11/38very hard · avg 14%▇ 1.7%at its level
AutomationBench v1.0.6 lab48.8% ✎▃····▃·▃▃··█ beats 4/6very hard · avg 5.7%▇ 1.3%at its level · small run
RobloxQA v2.0 multiple choice (Torpedo Software)94.6% ✎▁▁▃▁█▄▃▃▆▂▂▂ beats 28/49easy · avg 78%▇ 1.2%maxed out

$ where to run it

Through an API

api · $0.24/M · $0.15 in / $0.5 out per 1M tokens · 1M 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

  1. 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 ↗

  2. Run GLM-5.3-Flash · needs ~192.5 GB · from huggingface.co/unsloth
    chat in the terminal
    llama cli -hf unsloth/GLM-5.3-Flash-GGUF:UD-Q4_K_XL
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf unsloth/GLM-5.3-Flash-GGUF:UD-Q4_K_XL

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run GLM-5.3-Flash · needs ~192.5 GB · from huggingface.co/unsloth

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/unsloth/GLM-5.3-Flash-GGUF@UD-Q4_K_XL -y
  4. Then run lms ls to see its name, and lms chat followed by that name

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

$ how the score was worked out · Epoch AI — internal evaluatio… 54%, LMArena leaderboard dataset 26%, zai-org/GLM-5.3-Flash model c… 7% and 5 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

sourceresultsshare of the scorelicence
Epoch AI — internal evaluations6▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 54%CC BY 4.0
LMArena leaderboard dataset4▇▇▇▇▇▇▇▇ 26%CC BY 4.0
zai-org/GLM-5.3-Flash model card · lab-reported5▇▇ 7%MIT
Roblox LLM Leaderboard (Torpedo Software): RobloxQA v2.0 and Luau Leetcode2▇ 5%MIT
Terminal-Bench 4.0 official leaderboard submissions1▇ 4%Apache-2.0
Video2World: Benchmarking Coding Agents for Interactive World Modeling from Embodied Videos (arXiv:2610.04432v2)1▇ 2%CC BY-SA 4.0
Miðeind Icelandic LLM Leaderboard1▇ 2%CC BY 4.0
How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)1▇ 0%CC BY 4.0
prior · pull toward the middle▇ 0%
weighted average= 74.0

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit 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 · sources and licences · json