Zhipu AIGLM 5.3 Flash
GLM 5.3 Flash is a Zhipu AI open-weights reasoning model, released 26 Aug 2026. Its best configuration ranks #83 of 756 on the BenchLeader Index at 62.4 ±5.1, in the upper half. The ± is the point: 175 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in maths (69) and lowest in composite (59). At $0.238 per million tokens blended it is cheaper than most ranked models. Output speed of 54 tokens per second puts it slower than most, with a first answer in 40.2 s. It has been measured at 4 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 2 Oct 2026.
- Blended price
- $0.238/M
- $0.150 in · $0.500 out
- Output speed
- 54 tok/s
- measured by Artificial Analysis
- First answer
- 40 s
- first token 1.20 s
- Context
- 1M
- Full answer
- 49 s
- median, reasoning included
- Cost per run
- $0.253
- one full Intelligence Index run
- Released
- 26 Aug 2026
- Overall index62
- Reasoning67
- Coding64
- Agents & tools63
- Maths69
- Knowledge65
- Human preference67
- Multimodal65
- Long context65
- Composite59
Versions
Zhipu AI has shipped 3 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| GLM 5.3 Flashthis page | 26 Aug 2026 | 62.4 | #83 |
| GLM 4.7 Flash | 8 Aug 2025 | 51.6 | #328 |
| GLM-4.5-Flash | 28 Jul 2025 | – | – |
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| low | – | – | 54 tok/s | 40 s | $0.0002 | Reasoning 44 |
| high | – | – | 54 tok/s | 40 s | $0.0002 | Reasoning 55 |
| max | 54.3 | #239 | 54 tok/s | 40 s | $0.0002 | Agents & tools 51 · Coding 52 · Knowledge 59 · Maths 51 · Reasoning 64 |
| not statedbest | 62.4 | #83 | 54 tok/s | 40 s | $0.0002 | Agents & tools 63 · Coding 64 · Composite 59 · Human preference 67 · Knowledge 65 · Long context 65 · Maths 69 · Multimodal 65 · Reasoning 67 |
How its index has moved
8 Sept 2026 to 5 Oct 2026- not stated
- max
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Metadata
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| AA Openness Indexnot in index | – | – | – | 44.4#26 | Artificial Analysis |
Reasoning
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| GPQA Diamond | – | – | 90.2%#43 | – | Epoch AI Benchmarking Hub | |
| LMArena Hard Prompts | – | – | – | 1500#33 | LMArena | |
| LiveBench Reasoningnot in index | – | – | – | 77.6%#53 | LiveBench | |
| GPQA Diamond (AA)not in index | – | – | – | 91.2%#49 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | – | – | – | 39.9%#79 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | 86.4%#41 | – | Vals AI | |
| ARC-AGI-1 | 47.0%#153 | 71.8%#119 | 91.0%#60 | – | ARC Prize | |
| ARC-AGI-2 | 27.9%#116 | 50.1%#94 | 65.8%#64 | – | ARC Prize | |
| CritPt | – | – | – | 15.4%#79 | Artificial Analysis | |
| Chess Puzzlesnot in index | – | – | 14.0%#110 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | – | – | 8.0%#113 | – | Epoch AI Benchmarking Hub | |
| BTF-3not in index | – | – | – | 14.9%#1 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SciCode | – | – | – | 51.6%#82 | SciCode | |
| FrontierCode | – | – | 31.8%#27 | – | Cognition | |
| LMArena Coding | – | – | – | 1523#29 | LMArena | |
| LMArena WebDev | – | – | – | 1616#24 | LMArena | |
| LiveBench Codingnot in index | – | – | – | 79.0%#25 | LiveBench | |
| SciCode (AA)not in index | – | – | – | 51.6%#83 | Artificial Analysis | |
| LiveCodeBench | – | – | 80.5%#66 | – | Vals AI | |
| IOI | – | – | 52.5%#20 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | – | 92.0%#10 | – | Vals AI | |
| Code Migrationnot in index | – | – | 20.5%#43 | – | Vals AI | |
| ProgramBenchnot in index | – | – | 0.0%#23 | – | Vals AI | |
| Vibe Code Bench 1-100not in index | – | – | 16.0%#13 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | – | – | 30.8%#62 | – | Vals AI | |
| Surface Evolver Benchnot in index | – | – | 52.5%#17 | – | Epoch AI Benchmarking Hub | |
| DeepSWEnot in index | – | – | 63.4%#30 | – | Epoch AI Benchmarking Hub | |
| CursorBenchnot in index | 26.9%#57 | 31.1%#48 | 36.8%#34 | – | Cursor | |
| ALE-Benchnot in index | – | 303.6#110 | – | – | Epoch AI Benchmarking Hub | |
| FrontierSWEnot in index | – | – | 18.1%#15 | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | – | – | – | 32.8%#37 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | – | – | 84.3%#30 | Artificial Analysis |
Agents & tools
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| APEX-Agents | – | – | – | 52.8%#23 | Mercor | |
| LMArena Agent | – | – | – | -0.4#31 | LMArena | |
| LiveBench Agentic Codingnot in index | – | – | – | 56.8%#22 | LiveBench | |
| Terminal-Bench 2.1 (Vals) | – | – | 62.9%#38 | – | Vals AI | |
| GDPval (AA)not in index | – | – | – | 57.3%#16 | Artificial Analysis | |
| τ²-Bench Banking (AA) | – | – | – | 47.2%#8 | Artificial Analysis | |
| ITBench SRE (AA)not in index | – | – | – | 51.2%#4 | Artificial Analysis | |
| CyberBenchnot in index | – | – | 67.9%#23 | – | Vals AI | |
| Finance Agent v2not in index | – | – | 57.9%#12 | – | Vals AI | |
| Harvey's Legal Agent Benchmarknot in index | – | – | 6.7%#23 | – | Vals AI | |
| Legal Research Benchnot in index | – | – | 45.2%#14 | – | Vals AI | |
| SkillsBenchnot in index | – | – | 40.2%#29 | – | Vals AI | |
| Tax Agent Benchnot in index | – | – | 62.0%#28 | – | Vals AI | |
| Terminal-Bench Sciencenot in index | – | – | 2.9%#23 | – | Vals AI | |
| AutomationBenchnot in index | – | – | – | 60.4%#20 | Artificial Analysis | |
| GDP.pdfnot in index | – | – | – | 15.4%#35 | Artificial Analysis | |
| EnterpriseOps-Gymnot in index | – | – | – | 33.2%#14 | Artificial Analysis | |
| AA-Briefcasenot in index | – | – | – | 1454#31 | Artificial Analysis |
Maths
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | – | 55.8%#40 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | 17.1%#49 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | – | 93.9%#49 | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | 21.0%#47 | – | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | 81.2%#55 | LiveBench | |
| LMArena Maths | – | – | – | 1503#13 | LMArena |
Knowledge
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Data Analysisnot in index | – | – | – | 76.4%#32 | LiveBench | |
| AA-Omniscience | – | – | – | 7.5#93 | Artificial Analysis | |
| MMLU-Pro | – | – | 86.1%#47 | – | Vals AI | |
| LegalBench | – | – | 83.9%#43 | – | Vals AI | |
| TaxEval | – | – | 75.6%#16 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | – | – | 55.9%#41 | – | Vals AI | |
| MedScribenot in index | – | – | 88.9%#7 | – | Vals AI | |
| GDP.pdfnot in index | – | – | 14.0%#38 | – | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | – | – | – | 27.5%#187 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | – | – | – | 72.4%#31 | Artificial Analysis |
Instruction following
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | 77.3%#41 | LiveBench | |
| LiveBench Instruction Followingnot in index | – | – | – | 52.8%#63 | LiveBench | |
| LMArena Instruction Followingnot in index | – | – | – | 1472#35 | LMArena |
Human preference
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | 1473#41 | LMArena | |
| LMArena Creative Writingnot in index | – | – | – | 1433#68 | LMArena | |
| LMArena Multi-turnnot in index | – | – | – | 1471#58 | LMArena | |
| LMArena Longer Queriesnot in index | – | – | – | 1481#45 | LMArena |
Multimodal
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| LMArena Vision | – | – | – | 1301#18 | LMArena | |
| MMMU-Pro (Vals)not in index | – | – | 86.0%#23 | – | Vals AI |
Long context
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| AA-LCR | – | – | – | 80.0%#70 | Artificial Analysis | |
| MLCRnot in index | – | – | – | 51.1%#9 | Artificial Analysis |
Composite
| Benchmark | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | 151.9#42 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | – | 71.6%#50 | LiveBench | |
| AA Intelligence Index | – | – | – | 41.8#50 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Baseten | 98 tok/s | 0.76 s | $0.150 | $0.500 | 1.0M | fp8 |
| Fireworks | 97 tok/s | 0.57 s | $0.225 | $0.750 | 1.0M | – |
| Baseten | 83 tok/s | 0.65 s | $0.150 | $0.500 | 1.0M | fp8 |
| CoreWeave | 78 tok/s | 0.84 s | $0.150 | $0.500 | 1.0M | nvfp4 |
| Fireworks | 63 tok/s | 2.02 s | $0.150 | $0.500 | 1.0M | – |
| Crusoe | 62 tok/s | 1.07 s | $0.150 | $0.500 | 1.0M | fp4 |
| Together | 51 tok/s | 0.80 s | $0.150 | $0.500 | 1.0M | – |
| Relace | 41 tok/s | 1.54 s | $0.035 | $0.500 | 1.0M | – |
| Parasail | 40 tok/s | 1.24 s | $0.150 | $0.500 | 1.0M | fp4 |
| Modal | 37 tok/s | 0.47 s | $0.150 | $0.500 | 1.0M | nvfp4 |
| SiliconFlow | 33 tok/s | 1.79 s | $0.150 | $0.500 | 1.0M | fp8 |
| Wafer | 32 tok/s | 9.40 s | $0.060 | $0.500 | 1.0M | – |
| StreamLake | 31 tok/s | 2.63 s | $0.087 | $0.290 | 1.0M | fp8 |
| Decart | 31 tok/s | 3.22 s | $0.090 | $0.300 | 1.0M | fp4 |
| NextBit | 29 tok/s | 4.30 s | $0.165 | $0.550 | 1.0M | fp8 |
| Reka AI | 28 tok/s | 0.99 s | $0.150 | $0.500 | 262k | – |
| Z.ai | 28 tok/s | 4.37 s | $0.150 | $0.500 | 1.0M | fp8 |
| Cloudflare | 26 tok/s | 4.21 s | $0.300 | $1.00 | 1.0M | – |
| Friendli | 25 tok/s | 5.36 s | $0.150 | $0.500 | 1.0M | – |
| Phala | 24 tok/s | 5.28 s | $0.113 | $0.375 | 1.0M | fp8 |
| inference.net | 22 tok/s | 0.93 s | $0.044 | $0.600 | 1.0M | fp4 |
| AtlasCloud | 21 tok/s | 4.19 s | $0.150 | $0.500 | 1.0M | fp8 |
| DekaLLM | 20 tok/s | 1.24 s | $0.100 | $1.00 | 1.0M | – |
| GMICloud | 19 tok/s | 7.04 s | $0.090 | $0.300 | 1.0M | fp8 |
| Venice | 18 tok/s | 5.02 s | $0.150 | $0.500 | 1.0M | – |
| Inceptron | 18 tok/s | 5.17 s | $0.225 | $0.600 | 1.0M | fp8 |
| Open Inference | 17 tok/s | 3.99 s | $0.046 | $0.650 | 1.0M | fp4 |
| DeepInfra | 17 tok/s | 1.98 s | $0.075 | $0.250 | 1.0M | fp4 |
| NEAR AI | 17 tok/s | 3.12 s | $0.105 | $0.350 | 1.0M | fp8 |
| Sail Research | 16 tok/s | 1.48 s | $0.045 | $0.600 | 1.0M | fp4 |
| NovitaAI | 16 tok/s | 3.30 s | $0.084 | $0.280 | 1.0M | fp8 |
| DigitalOcean | 15 tok/s | 0.99 s | $0.150 | $0.500 | 1.0M | – |
| Morph | 10 tok/s | 5.46 s | $0.135 | $0.357 | 1.0M | fp8 |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.030 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0002 | $0.0002 | 45.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0021 | $0.0010 | 51.2 s |
| Code edit | 6,000 / 1,500 | $0.0017 | $0.0011 | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.011 | $0.0056 | 1.9 min |
| Structured extraction | 2,000 / 200 | $0.0004 | $0.0002 | 43.9 s |
See also
Data as of 5 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GLM 5.3 Flash: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/glm-5-3-flash, data as of 5 Oct 2026.