Zhipu AIGLM-5
GLM-5 is a Zhipu AI open-weights reasoning model, released 11 Feb 2026. Its best configuration (thinking reasoning effort) ranks #143 of 785 on the BenchLeader Index at 58.9 ±3.6, in the upper half. The ± is the point: 226 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 instruction following (71) and lowest in reasoning (52). At $1.55 per million tokens blended it is pricier than most ranked models. Output speed of 69 tokens per second puts it slower than most, with a first answer in 46.8 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 13 Oct 2026.
- Blended price
- $1.55/M
- $1.00 in · $3.20 out
- Output speed
- 69 tok/s
- measured by Artificial Analysis
- First answer
- 47 s
- first token 2.14 s
- Context
- 205k
- Full answer
- 54 s
- median, reasoning included
- Cost per run
- –
- Released
- 11 Feb 2026
- Overall index59
- Reasoning52
- Coding59
- Agents & tools58
- Maths63
- Knowledge58
- Instruction following71
- Long context63
- Composite61
Versions
Zhipu AI has shipped 8 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.
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 |
|---|---|---|---|---|---|---|
| no reasoning | 52.3 | #324 | 67 tok/s | 1.92 s | $0.0014 | Agents & tools 68 · Composite 54 · Instruction following 56 · Knowledge 56 · Long context 46 · Reasoning 41 |
| high | – | – | 57 tok/s | 2.14 s | $0.0014 | Coding 64 |
| thinkingbest | 58.9 | #143 | 69 tok/s | 47 s | $0.0014 | Agents & tools 58 · Coding 59 · Composite 61 · Instruction following 71 · Knowledge 58 · Long context 63 · Maths 63 · Reasoning 52 |
| not stated | 54.9 | #238 | 57 tok/s | 2.14 s | $0.0014 | Agents & tools 57 · Coding 52 · Human preference 64 · Maths 55 · Reasoning 51 |
How its index has moved
8 Sept 2026 to 23 Oct 2026- thinking
- not stated
- no reasoning
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
See all LLM benchmarks. 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 | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | – | 87.8%#70 | Epoch AI Benchmarking Hub | |
| SimpleBencheffort not stated | – | – | 53.2%#41 | SimpleBench | |
| Arena Hard Promptseffort not stated | – | – | 1478#67 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 66.6%#330 | 82.0%#175 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 7.7%#358 | 29.3%#157 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | 83.3%#56 | – | Vals AI | |
| Kagi LLM Benchmark | – | 75.0%#12 | 51.7%#80 | Kagi LLM Benchmark | |
| ARC-AGI-1effort not stated | – | – | 44.7%#186 | ARC Prize | |
| ARC-AGI-2effort not stated | – | – | 4.9%#177 | ARC Prize | |
| CritPt | 0.0%#315 | 2.0%#201 | – | Artificial Analysis | |
| CL-benchnot in indexeffort not stated | – | – | 18.7%#11 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | 63.7%#39 | ForecastBench |
Coding
| Benchmark | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SWE-bench Verified (Epoch)effort not stated | – | – | 72.1%#22 | Epoch AI Benchmarking Hub | |
| WeirdMLeffort not stated | – | – | 48.2%#73 | WeirdML | |
| Arena Codingeffort not stated | – | – | 1498#84 | Arena (LMArena) | |
| Arena WebDeveffort not stated | – | – | 1435#79 | Arena (LMArena) | |
| LiveCodeBench | – | 81.9%#57 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | 71.4%#55 | – | Vals AI | |
| SWE-bench Verified (bash only) | 72.8%#7 | – | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 72.8%#10 | – | – | SWE-bench | |
| Vibe Code Bench v1.1not in index | – | 23.4%#78 | – | Vals AI | |
| SWE Atlas: Codebase QnAnot in indexeffort not stated | – | – | 20.5%#18 | Scale AI SEAL | |
| SWE Atlas: Refactoringnot in indexeffort not stated | – | – | 24.2%#14 | Scale AI SEAL | |
| SWE Atlas: Test Writingnot in indexeffort not stated | – | – | 28.7%#19 | Scale AI SEAL | |
| ALE-Benchnot in indexeffort not stated | – | – | 765.6#79 | Epoch AI Benchmarking Hub |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | – | 52.4%#30 | Terminal-Bench | |
| Terminal-Bench Hard | 39.4%#57 | 43.2%#41 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 97.4%#12 | 98.3%#8 | – | Artificial Analysis | |
| τ²-bench | – | 81.0%#6 | – | τ²-bench | |
| τ³-bench Bankingnot in index | – | 9.8%#25 | – | τ²-bench | |
| APEX-Agents (AA)not in index | – | 14.4%#25 | – | Artificial Analysis | |
| Terminal-Bench 2.0 (Vals)not in index | – | 49.4%#21 | – | Vals AI | |
| Vending-Bench 2not in indexeffort not stated | – | – | 4432.1#38 | Andon Labs |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIMEeffort not stated | – | 80.0%#114 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 91.7%#23 | – | Vals AI | |
| AIME 2026effort not statedpossible contamination | – | 96.7%#6 | MathArena | |
| HMMT February 2026effort not stated | – | 86.4%#20 | MathArena | |
| USAMO 2026not in indexeffort not stated | – | 35.1%#9 | MathArena | |
| MathArena Apexeffort not statedpossible contamination | – | 10.9%#21 | MathArena | |
| Arena Mathseffort not stated | – | 1443#87 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -11.8#208 | 0.3#133 | Artificial Analysis | |
| MMLU-Pro | – | 86.0%#49 | Vals AI | |
| LegalBench | – | 84.1%#42 | Vals AI | |
| CorpFin | – | 62.9%#47 | Vals AI | |
| TaxEval | – | 70.0%#91 | Vals AI | |
| MedQA | – | 94.3%#18 | Vals AI | |
| CaseLaw v2not in index | – | 52.5%#44 | Vals AI | |
| AA-Omniscience: accuracynot in index | 23.1%#265 | 26.3%#217 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 54.7%#101 | 64.7%#64 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 55.2%#147 | 72.3%#49 | – | Artificial Analysis | |
| Arena Instruction Followingnot in indexeffort not stated | – | – | 1446#69 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1457#65 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1448#42 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1472#58 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1468#63 | Arena (LMArena) |
Long context
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 43.7%#344 | 75.7%#154 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 145.8#81 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 21.8#209 | 27.9#140 | – | 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 |
|---|---|---|---|---|---|---|
| Amazon Bedrock | 108 tok/s | 1.00 s | $1.00 | $3.20 | 203k | – |
| GMICloud | 66 tok/s | 2.38 s | $0.600 | $1.92 | 203k | fp8 |
| Baidu Qianfan | 61 tok/s | 0.92 s | $0.700 | $2.24 | 203k | fp8 |
| StreamLake | 60 tok/s | 1.96 s | $0.600 | $1.92 | 198k | fp8 |
| SiliconFlow | 49 tok/s | 2.37 s | $0.950 | $2.55 | 205k | fp8 |
| Venice (fp8) | 47 tok/s | 1.30 s | $1.00 | $3.20 | 198k | fp8 |
| Z.ai | 45 tok/s | 8.43 s | $1.00 | $3.20 | 203k | fp8 |
| NovitaAI | 43 tok/s | 1.89 s | $1.00 | $3.20 | 203k | fp8 |
| Venice | 36 tok/s | 8.97 s | $1.20 | $4.00 | 200k | – |
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 $0.200 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0014 | $0.0011 | 51.2 s |
| Summarize a 30-page report | 12,000 / 600 | $0.014 | $0.0067 | 55.6 s |
| Code edit | 6,000 / 1,500 | $0.011 | $0.0072 | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.073 | $0.037 | 1.8 min |
| Structured extraction | 2,000 / 200 | $0.0026 | $0.0014 | 49.7 s |
More from Zhipu AI
See also
Questions people ask
- What are GLM-5's benchmark scores?
- As of 24 Oct 2026, GLM-5 (thinking effort) scores 58.9 ±3.6 on the BenchLeader Index, strongest in instruction following (71) and weakest in reasoning (52). That rests on 15 independent benchmark results, and every score on this page links to its publisher.
- How does GLM-5 rank?
- GLM-5 (thinking effort) is #143 of 785 ranked configurations on 24 Oct 2026. 226 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does GLM-5 cost per 1M tokens?
- As of 24 Oct 2026, GLM-5 costs $1.00 per 1M input tokens and $3.20 per 1M output tokens, $1.55 blended (three input tokens per output token).
- When was GLM-5 released?
- Zhipu AI released GLM-5 on 11 Feb 2026. BenchLeader has results for it at 4 reasoning-effort settings.
Data as of 24 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GLM-5: benchmarks, pricing, speed and rank”, data as of 24 Oct 2026, https://www.benchleader.com/models/glm-5.