Zhipu AIGLM-4.6
GLM-4.6 is a Zhipu AI open-weights reasoning model, released 30 Sept 2025. Its best configuration ranks #384 of 785 on the BenchLeader Index at 49.9 ±4.5, in the upper half. The ± is the point: 320 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 human preference (60) and lowest in agents & tools (36). At $1.00 per million tokens blended it is mid-priced. Output speed of 26 tokens per second puts it in the slowest quarter, with a first token in 4.3 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 23 Oct 2026.
- Blended price
- $1.00/M
- $0.600 in · $2.20 out
- Output speed
- 26 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 4.25 s
- Context
- 205k
- Full answer
- –
- Cost per run
- –
- Released
- 30 Sept 2025
- Overall index50
- Reasoning52
- Coding45
- Agents & tools36
- Maths55
- Knowledge51
- Human preference60
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 | 46.7 | #479 | 35 tok/s | 5.08 s | $0.0009 | Agents & tools 59 · Composite 46 · Instruction following 40 · Knowledge 47 · Long context 37 · Reasoning 41 |
| thinking | 49.1 | #406 | 33 tok/s | 66 s | $0.0009 | Agents & tools 60 · Composite 50 · Instruction following 46 · Knowledge 43 · Long context 52 · Reasoning 43 |
| not statedbest | 49.9 | #384 | 26 tok/s | 4.25 s | $0.0009 | Agents & tools 36 · Coding 45 · Human preference 60 · Knowledge 51 · Maths 55 · Reasoning 52 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- not stated
- thinking
- 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 |
|---|---|---|---|---|---|
| Arena Hard Promptseffort not stated | – | – | 1442#120 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 63.2%#356 | 78.0%#214 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 5.5%#424 | 14.5%#257 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in indexeffort not stated | – | – | 74.5%#79 | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | 47.4%#90 | Kagi LLM Benchmark | |
| CritPt | 0.0%#315 | 1.1%#223 | – | Artificial Analysis |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| SciCodeeffort not stated | – | 38.4%#177 | SciCode | |
| Arena Codingeffort not stated | – | 1458#134 | Arena (LMArena) | |
| Arena WebDeveffort not stated | – | 1341#112 | Arena (LMArena) | |
| LiveCodeBencheffort not stated | – | 81.0%#62 | Vals AI | |
| SWE-Bench Proeffort not stated | – | 9.7%#22 | Scale AI SEAL | |
| SWE-bench Verified (bash only)effort not stated | – | 55.4%#28 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in indexeffort not stated | – | 68.2%#16 | SWE-bench | |
| Vibe Code Bench v1.1not in indexeffort not stated | – | 3.1%#107 | Vals AI | |
| SWE-Bench Pro (private)not in indexeffort not stated | – | 9.7%#22 | Scale AI SEAL | |
| ALE-Benchnot in indexeffort not stated | – | 340.8#113 | Epoch AI Benchmarking Hub | |
| Terminal-Bench 2.1 (AA)not in index | 49.4%#121 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | – | 24.5%#67 | Terminal-Bench | |
| Terminal-Bench Hard | 28.8%#117 | 25.0%#138 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 76.9%#133 | 70.5%#155 | – | Artificial Analysis | |
| BFCL Overall | – | 72.4%#4 | – | Berkeley Function Calling Leaderboard | |
| τ³-Banking (AA) | – | 13.4%#126 | – | Artificial Analysis | |
| Poker Agentnot in indexeffort not stated | – | – | 945.8%#15 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in indexeffort not stated | – | – | 28.1%#46 | Vals AI | |
| BrowseComp-Plusnot in indexeffort not stated | – | – | 27.7%#6 | BrowseComp-Plus |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AIME (Vals) | 92.7%#17 | Vals AI | |
| MGSM | 89.8%#41 | Vals AI | |
| MathArena Apexpossible contamination | 0.5%#41 | MathArena | |
| Arena Maths | 1418#123 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| AA-Omniscience | -31.7#288 | -41.9#332 | – | Artificial Analysis | |
| MMLU-Proeffort not stated | – | – | 82.2%#75 | Vals AI | |
| LegalBencheffort not stated | – | – | 79.6%#92 | Vals AI | |
| CorpFineffort not stated | – | – | 56.8%#86 | Vals AI | |
| TaxEvaleffort not stated | – | – | 66.2%#109 | Vals AI | |
| MedQAeffort not stated | – | – | 92.2%#33 | Vals AI | |
| AA-Omniscience: accuracynot in index | 21.4%#284 | 26.9%#210 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 32.4%#206 | 6.0%#511 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 36.7%#296 | 43.4%#223 | – | Artificial Analysis | |
| Arena Instruction Followingnot in indexeffort not stated | – | – | 1413#121 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1424#120 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1401#106 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1422#123 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1431#120 | Arena (LMArena) |
Long context
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 26.3%#407 | 54.0%#305 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | Source | Trend |
|---|---|---|---|---|
| AA Intelligence Index v4.3.2 | 14.9#289 | 18.5#252 | 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 |
|---|---|---|---|---|---|---|
| DeepInfra | 37 tok/s | 0.62 s | $0.500 | $2.00 | 203k | fp4 |
| Venice | 29 tok/s | 1.15 s | $0.430 | $1.75 | 198k | fp4 |
| Z.ai | 28 tok/s | 8.78 s | $0.600 | $2.20 | 203k | fp4 |
| NovitaAI | 26 tok/s | 5.21 s | $0.550 | $2.20 | 205k | bf16 |
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 $0.110 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0009 | $0.0008 | 15.8 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0085 | $0.0041 | 27.3 s |
| Code edit | 6,000 / 1,500 | $0.0069 | $0.0047 | 1.0 min |
| Agentic coding session | 60,000 / 4,000 | $0.045 | $0.023 | 2.6 min |
| Structured extraction | 2,000 / 200 | $0.0016 | $0.0009 | 11.9 s |
More from Zhipu AI
See also
Questions people ask
- What are GLM-4.6's benchmark scores?
- As of 25 Oct 2026, GLM-4.6 scores 49.9 ±4.5 on the BenchLeader Index, strongest in human preference (60) and weakest in agents & tools (36). That rests on 19 independent benchmark results, and every score on this page links to its publisher.
- How does GLM-4.6 rank?
- GLM-4.6 is #384 of 785 ranked configurations on 25 Oct 2026. 320 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-4.6 cost per 1M tokens?
- As of 25 Oct 2026, GLM-4.6 costs $0.600 per 1M input tokens and $2.20 per 1M output tokens, $1.00 blended (three input tokens per output token).
- When was GLM-4.6 released?
- Zhipu AI released GLM-4.6 on 30 Sept 2025. BenchLeader has results for it at 3 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “GLM-4.6: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/glm-4-6.