Models / GLM-5.3

Z.ai

GLM-5.3

GLM-5.3 is best for agents and tool use, and writing; a strong pick at a modest price. The cheapest model of similar intelligence is Claude Haiku 5.5, at $0.20 per million tokens against $2.15.

Results 16 Nov 2026 · 51 results · 7 sources

Test GLM-5.3 on your prompt → It runs beside the models you could use instead, with every answer checked and costed.

Price per million tokens
$1.40 in · $4.40 out
Z.ai's list price, 10 Nov 2026 · compare prices$0.26 cached input
Speed
87 tokens a second; on its test prompt the answer starts after 25.4 s, thinking time included
Artificial Analysis, on its usual API, at low reasoning; 78–87 across 2 settings
Intelligence Index
44.8
Artificial Analysis, at max reasoning; 34.3–44.8 across 2 settings
Spring Prompt overall
27th of 36 score 35 of 100 · how it works
Developer
Z.ai, based in China
Weights published
Released
14 Aug 2026
1,048,576 tokens of context

Where it stands among the models you could choose

Available models grouped by price, intelligence, speed and developer location, plus models with published weights. Retired models are excluded.

AmongIntelligence IndexSpring Prompt overallPrice, blendedSpeed
Models available today 17th of 185Claude Opus 5.5 27th of 35Claude Fable 5.2 $2.15 blended, 71st percentilecheapest: Mistral Nemo 40th of 65Nemotron 3.5 Super VL
Models priced $1–3 per million tokens 2nd of 34Muse Spark 1.3 4th of 8GPT-6.1 Terra sets this group 5th of 9Muse Spark 1.3
Models with similar intelligence sets this group 10th of 10GPT-6 Sol $2.15 blended, 40th percentilecheapest: Claude Haiku 5.5 7th of 12Muse Spark 1.3
Models writing 50–100 tokens a second 6th of 21Claude Opus 5.5 6th of 7Claude Opus 5.5 $2.15 blended, 70th percentilecheapest: Granite 4.2 8B sets this group
Models from developers based in China 4th of 75MiMo-V2.6-Pro 8th of 11Kimi K3.1 Pro $2.15 blended, 92nd percentilecheapest: Qwen3.5-9B 11th of 23DeepSeek-V4.1-Flash
Open-weights models 3rd of 104MiMo-V2.6-Pro 6th of 8Kimi K3.1 Pro $2.15 blended, 95th percentilecheapest: Mistral Nemo 24th of 43Nemotron 3.5 Super VL

The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.

Measured cost per task

API cost recorded in our benchmark runs, excluding judging.

TaskBenchmarkCost per task
Game of bullet chessBulletBench≈ $0.016

How good is it, and for what?

Each use case ranks the available models its benchmarks measured, then those priced $1–3 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.

Product listingsTurning a sparse product feed and photos into listings that can go live Not measured
  • CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is Not measured
  • DeckBench: has not measured this model.
User surveysPlanning a user survey and reading its results without being misled Not measured
  • SurveyBench: has not measured this model.
Marketing planningPlanning a year of ad spend without overspending 29th of 32 available · 6th of 7 at its price
BenchmarkGLM-5.3Rank among availableBest available
ROASBenchROASBench: overall score 24.8 29th of 32 Claude Fable 5.2 61.3
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 15th of 159 available · 3rd of 31 at its price
BenchmarkGLM-5.3Rank among availableBest available
Artificial AnalysisArtificial Analysis: Terminal-Bench 4.0 · reported 41.9% 13th of 107 Kimi K3.1 Pro 67.2%
Artificial AnalysisArtificial Analysis: Terminal-Bench 2.1 · reported 83.9% 16th of 107 Claude Fable 5.1 91.4%
Artificial AnalysisArtificial Analysis: τ-bench banking · reported 50.3% 3rd of 104 Grok 4.6 50.7%
Vending-Bench 2Vending-Bench 2: money after a year · reported $8,164 8th of 53 GPT-6 Astra $15,515
  • tau2-bench: pending. Its latest results are dated 26 May 2026, before this model was released.
  • Berkeley Function Calling Leaderboard (BFCL) V4: pending. Its latest results are dated 16 Dec 2025, before this model was released.
  • OpenHands Index: pending. Its latest results are dated 30 Jun 2026, before this model was released.
  • Microsoft STATE-Bench: pending. Its latest results are dated 29 May 2026, before this model was released.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 15th of 51 available · 2nd of 12 at its price
BenchmarkGLM-5.3Rank among availableBest available
APEX-AgentsAPEX-Agents: tasks passed · reported 56.6% 13th of 37 Claude Sonnet 5.5 75.5%
  • GDP.pdf: has not measured this model.
  • Remote Labor Index: has not measured this model.
Reasoning and knowledgeHard questions across science, maths and general knowledge 31st of 187 available · 3rd of 35 at its price
BenchmarkGLM-5.3Rank among availableBest available
Artificial AnalysisArtificial Analysis: Artificial Analysis Intelligence Index · reported 44.8 17th of 185 Claude Opus 5.5 57.6
Artificial AnalysisArtificial Analysis: Humanity's Last Exam · reported 42.3% 37th of 186 Claude Opus 5.5 61.4%
Artificial AnalysisArtificial Analysis: GPQA Diamond · reported 91.7% 24th of 168 GPT-6 Astra 96.3%
WritingWhat people prefer in blind comparisons, and judged writing quality 24th of 169 available · 4th of 28 at its price
BenchmarkGLM-5.3Rank among availableBest available
Arena (formerly LMArena)Arena Text: overall · reported 1,478 22nd of 153 Gemini 4 Argon 1,525
  • UGI Leaderboard: has not measured this model.
Sticking to the factsSummarising without inventing things, and factual answers 51st of 139 available · 7th of 26 at its price
BenchmarkGLM-5.3Rank among availableBest available
Arena (formerly LMArena)Arena Text factuality: overall · reported 1,467 25th of 109 Gemini 4 Argon 1,511
SimpleQA Verified (Epoch AI)SimpleQA Verified: correct answers · reported 41.0% 38th of 68 GPT-6 Astra 75.6%
  • Vectara Hallucination Leaderboard: has not measured this model.
Speed under pressureGood decisions against a real clock (fast chess) 33rd of 35 available · 6th of 7 at its price
BenchmarkGLM-5.3Rank among availableBest available
BulletBenchBulletBench Bullet 60s: ladder Elo 33 30th of 33 Gemini 3.8 Flash 1,068

Nearest alternatives

Against GLM-5.3 at $2.15 blended, Intelligence Index 44.8. "Similar intelligence" means within 4 points or better.

Where to run it

43 hosts on OpenRouter at the standard service tier, with a median of $1.50 blended per million tokens; quantisation where the host states it: FP4, FP8, NVFP4. Cheapest first.

HostQuantisationInputOutputBlended
WaferFP4$0.07$1.40$0.40
WaferFP4$0.18$1.40$0.48
Rekanot stated$0.14$2.10$0.63
DecartFP4$0.56$1.78$0.86
PrimeIntellectnot stated$0.40$2.46$0.92
InferenceNetFP4$0.08$3.60$0.96
InferenceNetFP4$0.08$3.60$0.96
NovitaFP8$0.63$1.98$0.97

Other service tiers, not counted above: Parasail fast $2.10 in, $6.60 out; Fireworks fast $2.10 in, $6.60 out; BaseTen fast $2.10 in, $6.60 out; BaseTen fast $2.10 in, $6.60 out. Flex and batch are cheaper but slower or queued; priority and fast cost more.

Prices per million tokens from OpenRouter's public endpoint list, checked 10 Nov 2026; the 35 dearer hosts are left out.

Against the alternatives

The models you would most likely weigh it against: the leaders of the groups above.

Scroll sideways to see every alternative.

GLM-5.3Muse Spark 1.3
most intelligent at $1–3 per million tokens
Claude Haiku 5.5
cheapest with similar intelligence
Claude Opus 5.5
most intelligent available today
MiMo-V2.6-Pro
most intelligent from China
Price per million tokens $2.15 $2.00$0.20$8.00$0.54
Tokens a second 87 2542409643
Intelligence Index 44.8 48.143.457.646.3
Spring Prompt overall 35 594082–
Rank among available models, by use case
Product listings – 17th20th13th–
Decks from an analysis – 25th18th7th–
User surveys – –7th4th–
Marketing planning 29th of 32 25th30th9th–
Agents and tool use 15th of 159 13th28th3rd23rd
Professional work 15th of 51 8th–3rd–
Reasoning and knowledge 31st of 187 11th25th1st13th
Writing 24th of 169 7th–5th23rd
Sticking to the facts 51st of 139 11th–3rd10th
Speed under pressure 33rd of 35 33rd21st––

Shaded figures are better than GLM-5.3's. A dash means no result.

Where it has been measured

2nd most intelligent of the 34 models priced $1–3 per million tokens, and 3rd most intelligent of the 104 open-weights models. Among models available today, in the top quarter (by percentile) for agents and tool use, writing, and reasoning and knowledge; in the bottom quarter for marketing planning and speed under pressure.

Every result

51 published results from 7 sources, each in the source's own units, with the configuration that produced it.

BulletBench · measured by Spring Prompt · 10 results
MeasureValueRankConfigurationDated
BulletBench Bullet 60s: cost per gameUS dollars, lower is better $0.0156 47 of 71 GLM-5.3
provider default reasoning
1 Oct 2026
BulletBench Bullet 60s: cost per gameUS dollars, lower is better $0.0220 55 of 71 GLM-5.3 (low reasoning)
low reasoning
1 Oct 2026
BulletBench Bullet 60s: games lost on time% of games, lower is better 91.7% 68 of 71 GLM-5.3
provider default reasoning
1 Oct 2026
BulletBench Bullet 60s: games lost on time% of games, lower is better 83.3% 60 of 71 GLM-5.3 (low reasoning)
low reasoning
1 Oct 2026
BulletBench Bullet 60s: invalid moves% of moves, lower is better 0.0% 1 of 71 GLM-5.3
provider default reasoning
1 Oct 2026
BulletBench Bullet 60s: invalid moves% of moves, lower is better 0.0% 1 of 71 GLM-5.3 (low reasoning)
low reasoning
1 Oct 2026
BulletBench Bullet 60s: ladder Eloladder Elo, higher is better 33
0–270
66 of 71 GLM-5.3
provider default reasoning
1 Oct 2026
BulletBench Bullet 60s: ladder Eloladder Elo, higher is better 33
0–270
66 of 71 GLM-5.3 (low reasoning)
low reasoning
1 Oct 2026
BulletBench Bullet 60s: median move timemilliseconds, lower is better 3.6 s 67 of 71 GLM-5.3
provider default reasoning
1 Oct 2026
BulletBench Bullet 60s: median move timemilliseconds, lower is better 1.6 s 43 of 71 GLM-5.3 (low reasoning)
low reasoning
1 Oct 2026

Not shown: Not your cost: prices are those charged on the run date.

ROASBench · measured by Spring Prompt · 9 results
MeasureValueRankConfigurationDated
ROASBench: audience scorescore out of 100, higher is better 51.0 28 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: business scorescore out of 100, higher is better 12.9 32 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: consistency scorescore out of 100, higher is better 14.8 30 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: contribution profitsimulated US dollars, higher is better $9,718 33 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: cost of a runUS dollars, lower is better $0.13 13 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: months over budgetmonths of 12, lower is better 0 1 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: overall scorescore out of 100, higher is better 24.8 32 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: planning scorescore out of 100, higher is better 51.7 35 of 35 GLM-5.3
provider default reasoning
29 Sep 2026
ROASBench: return on ad spendprofit per $1 spent, higher is better $0.01 33 of 35 GLM-5.3
provider default reasoning
29 Sep 2026

Not shown: Not real ad performance: the market is a deterministic simulation of one invented skincare brand.

APEX-Agents · reported by APEX-Agents · 1 result
MeasureValueRankConfigurationDated
APEX-Agents: tasks passed% of tasks, higher is better 56.6% 14 of 39 GLM-5.3 1 Oct 2026

Not shown: Not your firm's documents or tools; graded by rubric, not by a client.

Arena (formerly LMArena) · reported by Arena (formerly LMArena) · 11 results
MeasureValueRankConfigurationDated
Arena Agent: confirmed task successIPS effect estimate, higher is better 0.05
0.04–0.07
17 of 50 GLM-5.3 (max reasoning, arena agent) 2 Oct 2026
Arena Agent: praise over complaintIPS effect estimate, higher is better 0.05
0.03–0.08
19 of 50 GLM-5.3 (max reasoning, arena agent) 2 Oct 2026
Arena Agent: steerabilityIPS effect estimate, higher is better 0.01
-0.00–0.02
28 of 50 GLM-5.3 (max reasoning, arena agent) 2 Oct 2026
Arena Agent: tool groundingIPS effect estimate, higher is better 0.00
0.00–0.01
13 of 50 GLM-5.3 (max reasoning, arena agent) 2 Oct 2026
Arena Text factuality: overallArena rating, higher is better 1,467
1,462–1,472
34 of 182 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: business, management and financeArena rating, higher is better 1,475
1,464–1,486
31 of 406 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,454
1,444–1,464
34 of 411 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,518
1,504–1,532
19 of 363 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,476
1,467–1,484
25 of 413 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,478
1,473–1,484
27 of 413 GLM-5.3 (max reasoning) 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,466
1,457–1,475
29 of 412 GLM-5.3 (max reasoning) 2 Oct 2026

Not shown: Not a success rate: the estimated effect of choosing this model as the agent's orchestrator, against an even mix of the models tested, inside Arena's own harness.

Artificial Analysis · reported by Artificial Analysis · 18 results
MeasureValueRankConfigurationDated
Artificial Analysis: Artificial Analysis Coding Indexindex score, higher is better 74.8 29 of 187 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 34.3 91 of 413 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 44.8 37 of 413 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 91.7% 46 of 360 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 36.6% 110 of 419 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 42.3% 71 of 419 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 72.7% 177 of 407 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 79.7% 86 of 407 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Output speedtokens per second, higher is better 87.5 79 of 137 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Output speedtokens per second, higher is better 78.4 86 of 137 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: SciCode% of problems, higher is better 42.0% 156 of 205 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: SciCode% of problems, higher is better 59.0% 16 of 205 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 2.1% of tasks, higher is better 83.9% 34 of 188 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 4.0% of tasks, higher is better 34.8% 38 of 204 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 4.0% of tasks, higher is better 41.9% 30 of 204 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: Time to first answer tokenseconds, lower is better 25.4 90 of 137 GLM-5.3 (low reasoning) 20 Oct 2026
Artificial Analysis: Time to first answer tokenseconds, lower is better 28.2 96 of 137 GLM-5.3 (max reasoning) 20 Oct 2026
Artificial Analysis: τ-bench banking% of tasks, higher is better 50.3% 4 of 175 GLM-5.3 (max reasoning) 20 Oct 2026

Not shown: Not your codebase or tools.

SimpleQA Verified (Epoch AI) · reported by SimpleQA Verified (Epoch AI) · 1 result
MeasureValueRankConfigurationDated
SimpleQA Verified: correct answers% of questions, higher is better 41.0% 43 of 85 GLM-5.3 (max reasoning) 28 Aug 2026

Not shown: Not answers grounded in your documents; tests what the model remembers.

Vending-Bench 2 · reported by Vending-Bench 2 · 1 result
MeasureValueRankConfigurationDated
Vending-Bench 2: money after a yearUS dollars, higher is better $8,164 8 of 63 GLM-5.3 1 Oct 2026

Not shown: Not a real business; one simulated market with set rules.

Embed a score badge

A badge for each of our benchmarks that ranked GLM-5.3, from the current results. Paste the code into a page or README; it links back to the benchmark.

GLM-5.3 is #30 of 33 on ROASBench (Score: 24.8), measured by Spring Prompt ROASBench · Score
<a href="https://springprompt.com/benchmarks/roasbench?focus=zai-glm-5-3"><img src="https://springprompt.com/badge/roasbench/zai-glm-5-3.svg" alt="GLM-5.3 is #30 of 33 on ROASBench (Score: 24.8), measured by Spring Prompt" height="20"></a>
GLM-5.3 is #56 of 59 on BulletBench (Ladder Elo: 33), measured by Spring Prompt BulletBench · Ladder Elo
<a href="https://springprompt.com/benchmarks/bulletbench?focus=zai-glm-5-3"><img src="https://springprompt.com/badge/bulletbench/zai-glm-5-3.svg" alt="GLM-5.3 is #56 of 59 on BulletBench (Ladder Elo: 33), measured by Spring Prompt" height="20"></a>

Will GLM-5.3 work for your product?

Benchmarks aren't your task. Paste the prompt your product uses and run it beside the models you could use instead, with every answer checked and costed.

New here? Your email starts a free trial. No credit card needed.