Zhipu AIGLM-5.2
Best configuration ranks #138 of 610 on the BenchLeader Index at 57.6 ±3.7 (max reasoning effort). Last measured 8 Sept 2026. Released 16 Jun 2026.
- Blended price
- $1.79/M
- $1.10 in · $3.85 out
- Output speed
- 63 tok/s
- First answer
- 35 s
- first token 1.22 s
- Context
- 1M
- Overall index58
- Reasoning66
- Coding64
- Agents & tools57
- Maths56
- Knowledge45
- Human preference67
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. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 50.3 | #287 | 105 tok/s | 2.08 s | $0.0019 | Coding 39 · Composite 57 · Knowledge 61 · Long context 47 · Maths 46 · Reasoning 52 |
| low | – | – | 63 tok/s | 35 s | $0.0016 | Maths 58 · Reasoning 64 |
| high | – | – | 63 tok/s | 35 s | $0.0016 | Coding 62 |
| maxbest | 57.6 | #138 | 63 tok/s | 35 s | $0.0016 | Agents & tools 57 · Coding 64 · Human preference 67 · Knowledge 45 · Maths 56 · Reasoning 66 |
| thinking | – | – | 63 tok/s | 35 s | $0.0016 | Reasoning 55 |
| default | 57.4 | #139 | 63 tok/s | 35 s | $0.0019 | Agents & tools 63 · Coding 49 · Composite 62 · Human preference 54 · Instruction following 71 · Knowledge 61 · Long context 66 · Maths 47 · Reasoning 53 |
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.
Reasoning
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| GPQA Diamond | 71.2%#154 | 87.9%#53 | – | 91.9%#23 | – | – | Epoch AI Benchmarking Hub | |
| SimpleBench | – | – | – | – | – | 58.8%#32 | SimpleBench | |
| LMArena Hard Prompts | – | – | – | 1492#38 | – | – | LMArena | |
| LiveBench Reasoningnot in index | – | – | – | – | – | 78.6%#41 | LiveBench | |
| GPQA Diamond (AA)not in index | 68.6%#298 | – | – | – | – | 89.5%#74 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 9.8%#262 | – | – | – | – | 41.1%#49 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | – | – | – | 85.6%#44 | Vals AI | |
| Kagi LLM Benchmark | – | – | – | – | 60.0%#50 | 62.6%#44 | Kagi LLM Benchmark | |
| ARC-AGI-1 | – | – | – | – | – | 77.0%#87 | ARC Prize | |
| ARC-AGI-2 | – | – | – | – | – | 22.8%#93 | ARC Prize |
Coding
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SWE-bench Verified (Epoch) | – | – | – | 78.7%#5 | – | – | Epoch AI Benchmarking Hub | |
| SciCode | 36.1%#125 | – | – | 50.5%#56 | – | – | SciCode | |
| WeirdML | – | – | 67.3%#33 | 70.1%#30 | – | – | WeirdML | |
| FrontierCode | 24.5%#22 | – | – | – | – | – | Cognition | |
| LMArena Coding | – | – | – | 1509#47 | – | – | LMArena | |
| LMArena WebDev | – | – | – | 1589#16 | – | – | LMArena | |
| LiveBench Codingnot in index | – | – | – | – | – | 79.7%#14 | LiveBench | |
| SciCode (AA)not in index | – | – | – | – | – | 51.2%#61 | Artificial Analysis | |
| LiveCodeBench | – | – | – | – | – | 69.5%#89 | Vals AI | |
| SWE-bench (Vals)not in index | – | – | – | 82.8%#18 | – | – | Vals AI |
Agents & tools
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| APEX-Agents | – | – | – | – | – | 35.6%#17 | Mercor | |
| LMArena Agent | – | – | – | 4.6#11 | – | – | LMArena | |
| LiveBench Agentic Codingnot in index | – | – | – | – | – | 51.8%#28 | LiveBench | |
| Terminal-Bench Hard | – | – | – | – | – | 50.8%#22 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | – | – | – | – | – | 99.1%#1 | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | – | – | 67.8%#25 | – | – | Vals AI | |
| MCP Atlas | – | – | – | – | – | 77.8%#13 | Scale AI SEAL | |
| HiL-Bench | – | – | – | – | – | 43.7%#4 | Scale AI SEAL |
Maths
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | 42.5%#45 | 54.7%#36 | – | 59.2%#28 | – | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | – | – | 29.3%#28 | – | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | 28.9%#191 | 75.6%#111 | – | 86.4%#78 | – | – | Epoch AI Benchmarking Hub | |
| ProofBench | – | – | – | 35.0%#27 | – | – | Vals AI | |
| LiveBench Mathematicsnot in index | – | – | – | – | – | 89.8%#25 | LiveBench | |
| AIME 2026 | – | – | – | – | – | 90.0%#27 | MathArena | |
| HMMT February 2026 | – | – | – | – | – | 92.4%#13 | MathArena | |
| MathArena Apex | – | – | – | – | – | 19.8%#15 | MathArena |
Knowledge
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | – | 34.2%#45 | – | – | Epoch AI Benchmarking Hub | |
| LiveBench Data Analysisnot in index | – | – | – | – | – | 73.7%#31 | LiveBench | |
| AA-Omniscience | -6.6#119 | – | – | – | – | 4.4#75 | Artificial Analysis | |
| MMLU-Pro | – | – | – | – | – | 86.7%#38 | Vals AI | |
| LegalBench | – | – | – | – | – | 84.1%#36 | Vals AI | |
| CorpFin | – | – | – | – | – | 66.1%#21 | Vals AI | |
| TaxEval | – | – | – | – | – | 73.3%#49 | Vals AI |
Instruction following
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LiveBench Languagenot in index | – | – | – | – | – | 76.2%#35 | LiveBench | |
| IFBench | – | – | – | – | – | 73.3%#41 | Artificial Analysis |
Human preference
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| LMArena Text | – | – | – | 1472#36 | – | – | LMArena | |
| EQ-Bench 4 | – | – | – | – | – | 1222#13 | EQ-Bench |
Long context
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| AA-LCR | 42.3%#280 | – | – | – | – | 78.3%#77 | Artificial Analysis |
Composite
| Benchmark | no reasoning | low | high | max | thinking | default | Source | Trend |
|---|---|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | – | – | – | – | 151.9#37 | Epoch AI Benchmarking Hub | |
| LiveBench | – | – | – | – | – | 73.2%#36 | LiveBench | |
| AA Intelligence Index | 22.4#145 | – | – | – | – | 38.6#43 | Artificial Analysis | |
| Vals Indexnot in index | – | – | – | – | – | 53.1#23 | Vals AI |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
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 |
|---|---|---|---|---|---|---|
| Decart Fast | 225 tok/s | 1.06 s | $2.10 | $6.60 | 1.0M | fp4 |
| Fireworks Fast (US) | 136 tok/s | 0.91 s | $2.10 | $6.60 | 1.0M | – |
| Baseten Fast | 118 tok/s | 1.95 s | $2.10 | $6.60 | 1.0M | fp8 |
| Fireworks Fast | 106 tok/s | 0.72 s | $2.10 | $6.60 | 1.0M | – |
| Crusoe | 102 tok/s | 0.75 s | $1.40 | $4.40 | 1.0M | fp8 |
| Mistral | 99 tok/s | 0.73 s | $1.40 | $4.40 | 1.0M | – |
| Friendli | 99 tok/s | 0.43 s | $1.40 | $4.40 | 1.0M | – |
| Mistral (ZDR) | 97 tok/s | 1.01 s | $1.40 | $4.40 | 1.0M | – |
| Mistral (EU) | 80 tok/s | 1.22 s | $1.54 | $4.84 | 1.0M | – |
| Alibaba Fast | 80 tok/s | 0.92 s | $2.31 | $7.26 | 1.0M | fp8 |
| Baseten | 74 tok/s | 1.58 s | $1.40 | $4.40 | 1.0M | fp8 |
| AtlasCloud | 67 tok/s | 1.07 s | $0.938 | $2.95 | 1.0M | fp8 |
| Baidu Qianfan (fp4) | 65 tok/s | 0.92 s | $0.682 | $2.39 | 1.0M | fp4 |
| Baseten (US) | 61 tok/s | 2.23 s | $1.40 | $4.40 | 1.0M | fp8 |
| Baseten Fast (US) | 61 tok/s | 2.47 s | $2.10 | $6.60 | 1.0M | fp8 |
| Baidu Qianfan (fp8) | 60 tok/s | 0.91 s | $0.487 | $1.53 | 1.0M | fp8 |
| Alibaba Cloud Int. | 60 tok/s | 0.96 s | $0.966 | $3.04 | 1.0M | fp8 |
| CoreWeave | 58 tok/s | 1.46 s | $0.760 | $2.42 | 1.0M | fp4 |
| Venice | 56 tok/s | 1.89 s | $1.40 | $4.40 | 1M | fp8 |
| Phala | 51 tok/s | 1.54 s | $1.26 | $3.00 | 1.0M | fp8 |
| DigitalOcean | 48 tok/s | 0.91 s | $0.700 | $2.20 | 262k | – |
| SiliconFlow | 48 tok/s | 1.56 s | $1.19 | $3.74 | 1.0M | fp8 |
| Z.ai | 45 tok/s | 3.33 s | $1.40 | $4.40 | 1.0M | fp8 |
| Fireworks | 44 tok/s | 2.16 s | $1.40 | $4.40 | 1.0M | – |
| Together | 39 tok/s | 0.63 s | $1.40 | $4.40 | 512k | – |
| Parasail | 39 tok/s | 1.70 s | $1.40 | $4.40 | 262k | fp4 |
| StreamLake | 36 tok/s | 3.83 s | $0.280 | $0.880 | 1.0M | fp8 |
| DeepInfra | 36 tok/s | 1.03 s | $0.487 | $1.56 | 1.0M | fp4 |
| NovitaAI | 35 tok/s | 2.27 s | $0.683 | $2.15 | 1.0M | fp8 |
| Inceptron | 31 tok/s | 1.02 s | $1.25 | $2.99 | 1.0M | fp4 |
| Cloudflare | 23 tok/s | 4.24 s | $1.40 | $4.40 | 262k | – |
| GMICloud | 19 tok/s | 2.95 s | $1.40 | $4.40 | 1.0M | fp8 |
| Ambient | 13 tok/s | 2.91 s | $0.600 | $2.00 | 203k | 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 $0.275 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0016 | $0.0013 | 40.2 s |
| Summarise a 30-page report | 12,000 / 600 | $0.015 | $0.0081 | 45.0 s |
| Code edit | 6,000 / 1,500 | $0.012 | $0.0087 | 59.3 s |
| Agentic coding session | 60,000 / 4,000 | $0.081 | $0.044 | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.0030 | $0.0017 | 38.6 s |
See also
Data as of 9 Sept 2026. Compare these configurations.