GLM 5.2 vs Step 5 Preview
Full profiles: GLM 5.2 (Zhipu AI) and Step 5 Preview (StepFun): every benchmark, price and rank.
Updated · Data from 7 sources
As of 16 Nov 2026, GLM 5.2 and Step 5 Preview are neck and neck on quality (58.9 vs 56.0 on the BenchLeader Index, inside the ±5.9 margin), and Step 5 Preview is 1.5× cheaper at $1.43 vs $2.15 per 1M tokens blended. Choose GLM 5.2 for agents & tools (62 vs 49), coding (62 vs 50), and reasoning (57 vs 52); choose Step 5 Preview for knowledge (54 vs 44), maths (60 vs 53), and long context (69 vs 64), where it also streams ~1.5× faster.
Key facts (as of 16 Nov 2026)
- GLM 5.2 scores 58.9 on the BenchLeader Index and Step 5 Preview scores 56.0 (BenchLeader, 16 Nov 2026).
- GLM 5.2 costs $2.15 per 1M tokens blended (list price, 16 Nov 2026).
- Step 5 Preview costs $1.43 per 1M tokens blended (list price, 16 Nov 2026).
- Step 5 Preview streams ~1.5× faster than GLM 5.2 (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 35 benchmarks both models have, Step 5 Preview leads 22 (BenchLeader, 16 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | GLM 5.2 | 62 vs 50 category score; Arena Coding 1515 vs 1505 | High |
| Agents & tool use | GLM 5.2 | 62 vs 49 category score; Arena Agent 0.2 vs -3.7 | High |
| Maths | Step 5 Preview | 60 vs 53 category score; ProofBench 42.0% vs 35.0% | High |
| Reasoning | GLM 5.2 | 57 vs 52 category score; Arena Hard Prompts 1496 vs 1477 | Medium |
| Knowledge | Step 5 Preview | 54 vs 44 category score; AA-Omniscience v1.1 16.4 vs 0.6 | High |
| Long documents | Step 5 Preview | 69 vs 64 category score; AA-LCR v1.1 88.3% vs 78.3% | Medium |
| Chat quality (human preference) | GLM 5.2 | 66 vs 63 on Arena votes | Medium |
| Lowest cost | Step 5 Preview | $1.43 vs $2.15 per 1M blended; agentic coding session $0.071 vs $0.102 | – |
| Fastest first answer | Either | 26.8 s (GLM 5.2 (max)) vs 23.5 s (Step 5 Preview) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- GLM 5.2 (max) and Step 5 Preview are neck and neck on quality (58.9 vs 56.0, within the ±5.9 margin).
- GLM 5.2 (max) is stronger in agents & tools, coding, human preference, reasoning.
- Step 5 Preview is stronger in knowledge, long context, maths.
- Step 5 Preview is 1.5× cheaper ($1.43 vs $2.15 per 1M blended).
- Step 5 Preview streams 1.8× faster (124 vs 69 tokens per second).
At a glance
| Spec | GLM 5.2 | Step 5 Preview |
|---|---|---|
| Developer | Zhipu AI | StepFun |
| Released | 16 Jun 2026 | 18 Sep 2026 |
| Context window (tokens) | 1,000,000 | 1,000,000 |
| Compared at | GLM 5.2 (max) | Step 5 Preview |
| Open weights | Yes | Yes |
| Price in / out / cached input per 1M | $1.40 / $4.40 / $0.260 | $1.00 / $2.70 / $0.105 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | GLM 5.2 (max) | Step 5 Preview | Leader |
|---|---|---|---|
| Reasoning | 57.2 | 51.7 | GLM 5.2 |
| Coding | 61.8 | 49.9 | GLM 5.2 |
| Agents & tools | 62.1 | 49.1 | GLM 5.2 |
| Maths | 52.6 | 60.1 | Step 5 Preview |
| Knowledge | 44.3 | 53.7 | Step 5 Preview |
| Instruction following | 71.8 | – | not comparable |
| Human preference | 66.4 | 63.1 | GLM 5.2 |
| Multimodal | – | 59.3 | not comparable |
| Long context | 63.8 | 69.0 | Step 5 Preview |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Instruction following, Multimodal.
Where sources disagree
Across the 35 benchmarks both have, GLM 5.2 leads 12, Step 5 Preview leads 22 and 1 is a tie. Every overall index that rates both puts GLM 5.2 ahead.
| Overall index | GLM 5.2 (max) | Step 5 Preview | Leader |
|---|---|---|---|
| BenchLeader Index | 58.9 | 56.0 | neck and neck |
| Arena Text | 1475 | 1449 | GLM 5.2 |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | GLM 5.2 (max) | Step 5 Preview |
|---|---|---|
| Blended price per 1M tokens | $2.15 | $1.43 |
| Input / output per 1M | $1.40 / $4.40 | $1.00 / $2.70 |
| Cached input per 1M | $0.260 | $0.105 |
| Price per task: chat reply | $0.0019 | $0.0012 |
| Price per task: summarize a 30-page report | $0.019 | $0.014 |
| Price per task: code edit | $0.015 | $0.010 |
| Price per task: agentic coding session | $0.102 | $0.071 |
| Price per task: structured extraction | $0.0037 | $0.0025 |
| Agentic coding session, input mostly cached | $0.050 | $0.030 |
| Cost per benchmark task (AA Index) | – | $2.35 |
| Output tokens per task (AA Index) | – | 113k |
| Of which reasoning | – | 80k |
| Time per task (AA Index) | – | 15 min |
| Output speed | 69 tok/s | 124 tok/s |
| Time to first answer | 26.8 s | 23.5 s |
| Time to a complete answer, last measured 13 Nov 2026 | 56.7 s | 54.4 s |
Each model is compared at its best-ranked reasoning-effort setting. GLM 5.2 (max) scores 58.9 and takes 27 s to its first answer token; GLM 5.2 scores 51.9 in 1 s, GLM 5.2 (no reasoning) scores 45.8 in 2 s. Step 5 Preview scores 56.0 and takes 24 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
150 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | GLM 5.2 (max) | Step 5 Preview |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 58.9 | 56.0 |
| Agents & tools score | BenchLeader | 62.1 | 49.1 |
| Coding score | BenchLeader | 61.8 | 49.9 |
| Human preference score | BenchLeader | 66.4 | 63.1 |
| Instruction following score | BenchLeader | 71.8 | – |
| Knowledge score | BenchLeader | 44.3 | 53.7 |
| Long context score | BenchLeader | 63.8 | 69.0 |
| Maths score | BenchLeader | 52.6 | 60.1 |
| Reasoning score | BenchLeader | 57.2 | 51.7 |
| Multimodal score | BenchLeader | – | 59.3 |
| Blended price $/M | BenchLeader | $2.15 | $1.43 |
| Output speed | BenchLeader | 69 tok/s | 124 tok/s |
| Time to first answer | BenchLeader | 26.8 s | 23.5 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,000,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 91.9% | – |
| SimpleBench | SimpleBench | 58.8%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1496 | 1477 |
| LiveBench 2.0 Reasoning | LiveBench | – | 83.8%† |
| LiveBench 1.0 Reasoning | LiveBench | 78.6%† | – |
| GPQA Diamond (AA) | Artificial Analysis | 89.5% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 41.1% | 46.5% |
| HLE-Diamond (AA) | Artificial Analysis | 17.7% | 24.3% |
| GPQA Diamond (Vals) | Vals AI | 85.6%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 62.6%† | – |
| ARC-AGI-1 | ARC Prize | 77.0%† | – |
| ARC-AGI-2 | ARC Prize | 22.8%† | – |
| CritPt | Artificial Analysis | 20.9% | 20.9% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 32.5% |
| MysteryMechanism | Vals AI | – | 18.5% |
| Arena Expert | Arena (LMArena) | 1502 | 1489 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 22.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 18.0%† | – |
| LMCA | Epoch AI Benchmarking Hub | 45.8% | – |
| DTBench | Epoch AI Benchmarking Hub | 93.6% | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 78.7% | – |
| SciCode | SciCode | 51.2% | 58.9% |
| WeirdML | WeirdML | 70.1% | – |
| FrontierCode | Cognition | 24.5%† | – |
| Arena Coding | Arena (LMArena) | 1515 | 1505 |
Show all 150 benchmarks
| Metric | Publisher | GLM 5.2 (max) | Step 5 Preview |
|---|---|---|---|
| Coding | |||
| Arena WebDev | Arena (LMArena) | 1608 | 1573† |
| LiveBench 2.0 Coding | LiveBench | – | 67.6%† |
| LiveBench 1.0 Coding | LiveBench | 79.7%† | – |
| LiveCodeBench | Vals AI | 69.5%† | – |
| IOI | Vals AI | – | 38.1% |
| SWE-bench (Vals) | Vals AI | 82.8% | – |
| Code Migration | Vals AI | 37.9% | 39.0% |
| Code Migration v1.1 | Vals AI | – | 47.0% |
| ProgramBench | Vals AI | 0.5% | – |
| SWE-bench Multimodal | Vals AI | – | 40.8% |
| Vibe Code Bench v1.1 | Vals AI | 64.0% | 70.4% |
| SWE Atlas v1.0: Codebase QnA | Scale AI SEAL | 48.1%† | – |
| SWE Atlas v1.0: Refactoring | Scale AI SEAL | 42.4%† | – |
| SWE Atlas v1.0: Test Writing | Scale AI SEAL | 41.5%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 36.7% | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 55.6%† | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 43.8% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1010.2 | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 1.0% | 33.3% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 77.9% | – |
| Startrise Frontend | Startrise | 72.1%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 43.1† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 32.4† | – |
| Design Arena Website | Design Arena | 1277† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 38.4% |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 12.0% |
| Agents & tools | |||
| APEX-Agents | Mercor | 45.2%† | – |
| Arena Agent | Arena (LMArena) | 0.2 | -3.7 |
| LiveBench Long Agentic | LiveBench | – | 19.4%† |
| LiveBench Browser | LiveBench | – | 24.8%† |
| LiveBench Agentic Coding | LiveBench | 51.8%† | – |
| Terminal-Bench Hard | Artificial Analysis | 50.8% | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 99.1% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 67.8% | – |
| MCP Atlas | Scale AI SEAL | 77.8%† | – |
| HiL-Bench | Scale AI SEAL | 43.7%† | – |
| τ³-bench Banking | τ²-bench | 37.1%† | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 54.3% |
| GDPval-AA v2.1 | Artificial Analysis | 1380 | 1591 |
| τ³-Banking (AA) | Artificial Analysis | 34.6% | – |
| ITBench SRE (AA) | Artificial Analysis | 42.7% | 55.6% |
| Analyst Agent (AA) | Artificial Analysis | – | 35.0% |
| APEX-Agents (AA) | Artificial Analysis | 33.7% | 38.0% |
| Finance Agent v2 | Vals AI | 49.7%† | 50.7% |
| Finance Agent v2.1 | Vals AI | – | 58.9% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 12.9% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 7.1% | 9.2% |
| Legal Research Bench | Vals AI | 31.3% | 40.4% |
| SkillsBench | Vals AI | 45.1% | – |
| SREBench | Vals AI | 0.0% | 0.4% |
| Tax Agent Bench | Vals AI | – | 27.3% |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 31.8% |
| Terminal-Bench Science | Vals AI | – | 2.9% |
| Arena Agent Task Completion | Arena (LMArena) | 1.7 | -1.5 |
| PostTrainBench | Epoch AI Benchmarking Hub | 30.4% | – |
| GBAEval | Epoch AI Benchmarking Hub | 0.0%† | – |
| Vending-Bench 2 | Andon Labs | 8313.8† | – |
| AutomationBench | Artificial Analysis | 28.4% | 51.0% |
| GDP.pdf (AA) | Artificial Analysis | 10.4% | 14.8% |
| EnterpriseOps-Gym | Artificial Analysis | 42.7% | 47.2% |
| AA-Briefcase v1.1 | Artificial Analysis | 1226 | 1424 |
| Design Arena Agentic | Design Arena | 1194† | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 2.4% |
| CWE-Bench (AA) | Artificial Analysis | – | 52.5% |
| DeepSec-Bench (AA) | Artificial Analysis | – | 15.8% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 74.8% |
| MKA (AA) | Artificial Analysis | – | 27.6% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 59.2% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 29.3% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 86.4% | – |
| ProofBench | Vals AI | 35.0% | 42.0% |
| LiveBench Mathematics | LiveBench | 89.8%† | – |
| AIME 2026 | MathArena | 90.0% | – |
| HMMT February 2026 | MathArena | 92.4% | – |
| MathArena Apex | MathArena | 19.8% | – |
| Arena Maths | Arena (LMArena) | 1487 | 1470 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 34.2% | – |
| LiveBench 2.0 Data Analysis | LiveBench | – | 63.5%† |
| LiveBench Research | LiveBench | – | 75.5%† |
| LiveBench 1.0 Data Analysis | LiveBench | 73.7%† | – |
| AA-Omniscience v1.1 | Artificial Analysis | 0.6 | 16.4 |
| MMLU-Pro | Vals AI | 86.7%† | – |
| LegalBench | Vals AI | 84.1%† | – |
| CorpFin | Vals AI | 66.1%† | – |
| TaxEval | Vals AI | 73.3%† | – |
| BioMysteryBench | Vals AI | – | 67.4% |
| Excel Modeling Benchmark | Vals AI | 61.5%† | 57.1% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 69.5% |
| MedCode | Vals AI | 40.8%† | – |
| MedScribe | Vals AI | 83.5%† | – |
| Public Benefits Bench v1.1 | Vals AI | – | 58.1% |
| DrugDiscoveryBench | Scale AI SEAL | 36.2%† | – |
| EBR-bench | Epoch AI Benchmarking Hub | 9.5% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 21.1% | 36.4% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 74.1% | 68.7% |
| Instruction following | |||
| LiveBench Language | LiveBench | 76.2%† | – |
| LiveBench Instruction Following | LiveBench | 62.3%† | – |
| IFBench | Artificial Analysis | 73.3% | – |
| Arena Instruction Following | Arena (LMArena) | 1470 | 1450 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1475 | 1449 |
| EQ-Bench 4 | EQ-Bench | 1208† | – |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1759† | – |
| Arena Creative Writing | Arena (LMArena) | 1451 | 1413 |
| Arena Multi-turn | Arena (LMArena) | 1477 | 1445 |
| Arena Longer Queries | Arena (LMArena) | 1486 | 1461 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | – | 1263 |
| LiveBench Design | LiveBench | – | 70.8%† |
| MMMU-Pro | Artificial Analysis | – | 76.4% |
| Code2Video-AA | Artificial Analysis | – | 821 |
| AA-3D | Artificial Analysis | – | 859 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 78.3% | 88.3% |
| MLCR | Artificial Analysis | 12.8% | 18.3% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 153.0† | – |
| LiveBench 2.0 | LiveBench | – | 54.9%† |
| LiveBench 1.0 | LiveBench | 73.2%† | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 28.5 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 33.7 | 43.7 |
| AA Progress Index v1 | Artificial Analysis | 302.5 | 318.2 |
| Vals Index | Vals AI | – | 54.8 |
| Vals Finance Index | Vals AI | – | 55.4% |
| Vals Science Index | Vals AI | – | 29.6% |
| Vals Legal Index | Vals AI | – | 70.2% |
| AA Cyber Index | Artificial Analysis | – | 47.7 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 48.7%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 44.4 | 33.3 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Step 5 Preview: BenchLeader Index 54.2 → 56.0.
- Step 5 Preview: BenchLeader Index 54.7 → 54.2.
- GLM 5.2 (max): 1 new result (GDPval-AA v2.1).
- GLM 5.2 (max): BenchLeader Index 59.5 → 59.0.
- GLM 5.2 (max): 3 new results (AIME 2026, HMMT February 2026, MathArena Apex); 1 score revision.
- Step 5 Preview: 1 score revision.
- GLM 5.2 (max): 1 new result (MirrorCode).
- Step 5 Preview: 7 score revisions.
- GLM 5.2 (max): BenchLeader Index 60.2 → 59.5.
- GLM 5.2 (max): 4 new results (HLE-Diamond (AA), AA-Omniscience v1.1, AA-Omniscience v1.1: accuracy…).
Related comparisons
- GLM 5.2 vs Claude Opus 5.5
- Step 5 Preview vs Claude Opus 5.5
- Step 5 Preview vs GPT-6 Astra
- Step 5 Preview vs GPT-6.1 Sol
- Step 5 Preview vs Gemini 4 Argon
- GLM 5.2 vs GPT-6.1 Sol
- GLM 5.2 vs GPT-5.6 Sol
- GLM 5.2 vs Gemini 4 Argon
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GLM 5.2 vs Step 5 Preview: questions
- Is GLM 5.2 better than Step 5 Preview?
- GLM 5.2 (max) and Step 5 Preview are neck and neck on quality (58.9 vs 56.0, within the ±5.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- GLM 5.2 vs Step 5 Preview: which is cheaper, and will that change?
- Step 5 Preview is cheaper today: $1.43 against $2.15 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is GLM 5.2 or Step 5 Preview better for coding?
- GLM 5.2 scores higher in coding (62 vs 50 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 51.2% for GLM 5.2 against 58.9% for Step 5 Preview.
- Which is faster, GLM 5.2 or Step 5 Preview?
- Step 5 Preview streams ~1.5× faster (124 vs 69 output tokens per second, Artificial Analysis 7-day median).
- Is GLM 5.2 or Step 5 Preview better for agents and tool use?
- GLM 5.2 scores higher in agentic tasks (62 vs 49 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is 0.2 for GLM 5.2 against -3.7 for Step 5 Preview.
More questions (4)
- Which is better for long documents?
- Step 5 Preview scores higher on long-context benchmarks (69 vs 64). Both accept 1,000,000 tokens of context.
- Which has lower latency?
- Step 5 Preview answers first: 23.5 s against 26.8 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on GLM 5.2 vs Step 5 Preview?
- Across the 35 benchmarks both have, GLM 5.2 leads 12, Step 5 Preview leads 22 and 1 is a tie. Every overall index that rates both puts GLM 5.2 ahead.
- When was Step 5 Preview released?
- Step 5 Preview was released on 18 Sep 2026.