GLM-5 vs o3
Full profiles: GLM-5 (Zhipu AI) and o3 (OpenAI): every benchmark, price and rank.
Updated · Data from 15 sources
As of 29 Oct 2026, GLM-5 and o3 are neck and neck on quality (60.2 vs 61.0 on the BenchLeader Index, inside the ±4.9 margin), and GLM-5 is 2.3× cheaper at $1.55 vs $3.50 per 1M tokens blended. Choose GLM-5 for reasoning (68 vs 61), instruction following (71 vs 65), and knowledge (57 vs 56); choose o3 for fast first answers (5.6 s vs 46.9 s) and agents & tools (70 vs 58), maths (71 vs 63), and coding (62 vs 59), where it also streams ~2× faster.
Key facts (as of 29 Oct 2026)
- GLM-5 scores 60.2 on the BenchLeader Index and o3 scores 61.0 (BenchLeader, 29 Oct 2026).
- GLM-5 costs $1.55 per 1M tokens blended (list price, 29 Oct 2026).
- o3 costs $3.50 per 1M tokens blended (list price, 29 Oct 2026).
- o3 streams ~2× faster than GLM-5 (Artificial Analysis, 7-day median, 29 Oct 2026).
- Of the 9 benchmarks both models have, GLM-5 leads 8 (BenchLeader, 29 Oct 2026).
GLM-5 benchmarks 
o3 benchmarks 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 | o3 | 62 vs 59 category score | Medium |
| Agents & tool use | o3 | 70 vs 58 category score; Terminal-Bench Hard 37.1% vs 43.2% | High |
| Maths | o3 | 71 vs 63 category score | Medium |
| Reasoning | GLM-5 | 68 vs 61 category score; Kagi LLM Benchmark 75.0% vs 67.6% | Medium |
| Knowledge | GLM-5 | 57 vs 56 category score | Medium |
| Long documents | Either | 63 vs 62 category score; AA-LCR v1.1 75.7% vs 74.7% | Either |
| Following instructions | GLM-5 | 71 vs 65 category score; IFBench 72.3% vs 71.4% | Medium |
| Chat quality (human preference) | – | Only o3 has an Arena rating (61 category score) | Not comparable |
| Lowest cost | GLM-5 | $1.55 vs $3.50 per 1M blended; agentic coding session $0.073 vs $0.152 | – |
| Fastest first answer | o3 | 5.6 s (o3) vs 46.9 s (GLM-5 (thinking)) 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 (thinking) and o3 are neck and neck on quality (60.2 vs 61.0, within the ±4.9 margin).
- GLM-5 (thinking) is stronger in instruction following, knowledge, long context, reasoning.
- o3 is stronger in agents & tools, coding, maths, human preference, multimodal.
- GLM-5 (thinking) is 2.3× cheaper ($1.55 vs $3.50 per 1M blended).
- o3 streams 2.2× faster (151 vs 69 tokens per second).
At a glance
| Spec | GLM-5 | o3 |
|---|---|---|
| Developer | Zhipu AI | OpenAI |
| Released | 11 Feb 2026 | 16 Apr 2025 |
| Context window (tokens) | 204,800 | 200,000 |
| Compared at | GLM-5 (thinking) | o3 |
| Open weights | Yes | No |
| Price in / out / cached input per 1M | $1.00 / $3.20 / $0.200 | $2.00 / $8.00 / $0.500 |
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 (thinking) | o3 | Leader |
|---|---|---|---|
| Reasoning | 68.2 | 60.9 | GLM-5 |
| Coding | 59.1 | 61.8 | o3 |
| Agents & tools | 58.0 | 70.3 | o3 |
| Maths | 63.2 | 70.5 | o3 |
| Knowledge | 57.3 | 56.1 | GLM-5 |
| Instruction following | 70.8 | 65.1 | GLM-5 |
| Human preference | – | 61.1 | not comparable |
| Multimodal | – | 56.7 | not comparable |
| Long context | 62.6 | 62.1 | level |
Where sources disagree
Across the 9 benchmarks both have, GLM-5 leads 8, o3 leads 1. Every overall index that rates both puts GLM-5 ahead.
| Overall index | GLM-5 (thinking) | o3 | Leader |
|---|---|---|---|
| BenchLeader Index | 60.2 | 61.0 | neck and neck |
| Arena Text | 1457† | 1432 | GLM-5 |
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; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | GLM-5 (thinking) | o3 |
|---|---|---|
| Blended price per 1M tokens | $1.55 | $3.50 |
| Input / output per 1M | $1.00 / $3.20 | $2.00 / $8.00 |
| Cached input per 1M | $0.200 | $0.500 |
| Price per task: chat reply | $0.0014 | $0.0032 |
| Price per task: summarize a 30-page report | $0.014 | $0.029 |
| Price per task: code edit | $0.011 | $0.024 |
| Price per task: agentic coding session | $0.073 | $0.152 |
| Price per task: structured extraction | $0.0026 | $0.0056 |
| Output speed | 69 tok/s | 151 tok/s |
| Time to first answer | 46.9 s | 5.6 s |
| Time to a complete answer | 54.2 s | 8.9 s |
Each model is compared at its best-ranked reasoning-effort setting. GLM-5 (thinking) scores 60.2 and takes 47 s to its first answer token; GLM-5 scores 54.9 in 2 s. o3 scores 61.0 and takes 6 s to its first answer token; o3 (high) scores 53.8 in 6 s, o3 (medium) scores 53.3 in 6 s. 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
106 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 (thinking) | o3 |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 60.2 | 61.0 |
| Agents & tools score | BenchLeader | 58.0 | 70.3 |
| Coding score | BenchLeader | 59.1 | 61.8 |
| Instruction following score | BenchLeader | 70.8 | 65.1 |
| Knowledge score | BenchLeader | 57.3 | 56.1 |
| Long context score | BenchLeader | 62.6 | 62.1 |
| Maths score | BenchLeader | 63.2 | 70.5 |
| Reasoning score | BenchLeader | 68.2 | 60.9 |
| Human preference score | BenchLeader | – | 61.1 |
| Multimodal score | BenchLeader | – | 56.7 |
| Blended price $/M | BenchLeader | $1.55 | $3.50 |
| Output speed | BenchLeader | 69 tok/s | 151 tok/s |
| Time to first answer | BenchLeader | 46.9 s | 5.6 s |
| Context window | BenchLeader | 204,800 tokens | 200,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 87.8%† | 81.8%† |
| Humanity's Last Exam | Scale AI / CAIS | – | 20.3%† |
| SimpleBench | SimpleBench | 53.2%† | 53.1%† |
| Arena Hard Prompts | Arena (LMArena) | 1478† | 1441 |
| GPQA Diamond (AA) | Artificial Analysis | 82.0% | 82.7% |
| Humanity's Last Exam (AA) | Artificial Analysis | 29.3% | 20.1% |
| GPQA Diamond (Vals) | Vals AI | 83.3% | 84.1%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 75.0% | 67.6% |
| GPQA Diamond (HELM) | HELM Capabilities | – | 75.3% |
| ARC-AGI-1 | ARC Prize | 44.7%† | 60.8%† |
| ARC-AGI-2 | ARC Prize | 4.9%† | 6.5%† |
| CritPt | Artificial Analysis | 2.0% | 1.1% |
| SciPredict | Scale AI SEAL | – | 17.9% |
| Chess Puzzles | Epoch AI Benchmarking Hub | – | 36.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | – | 29.0%† |
| CL-bench | Epoch AI Benchmarking Hub | 18.7%† | 17.8%† |
| LMCA | Epoch AI Benchmarking Hub | – | 39.7%† |
| DTBench | Epoch AI Benchmarking Hub | – | 84.8%† |
| ForecastBench | ForecastBench | 63.7%† | 64.6% |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 72.1%† | 62.3%† |
| WeirdML | WeirdML | 48.2%† | 52.4%† |
| GSO-Bench | GSO-Bench | – | 8.8%† |
| Arena Coding | Arena (LMArena) | 1498† | 1460 |
| Arena WebDev | Arena (LMArena) | 1436† | – |
| LiveCodeBench | Vals AI | 81.9% | 83.9%† |
Show all 106 benchmarks
| Metric | Publisher | GLM-5 (thinking) | o3 |
|---|---|---|---|
| Coding | |||
| SWE-bench (Vals) | Vals AI | 71.4% | – |
| Aider Polyglot | Aider polyglot leaderboard | – | 81.3% |
| SWE-bench Verified (bash only) | SWE-bench | 72.8%† | 58.4% |
| SWE-bench Verified (any scaffold) | SWE-bench | 72.8%† | 58.4% |
| Vibe Code Bench v1.1 | Vals AI | 23.4% | – |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 20.5%† | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 24.2%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 28.7%† | – |
| CadEval | Epoch AI Benchmarking Hub | – | 74.0%† |
| ALE-Bench | Epoch AI Benchmarking Hub | 765.6† | 933.5† |
| LiveCodeBench (AA) | Artificial Analysis | – | 80.8% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 52.4%† | – |
| GDPval | OpenAI | – | 30.8%† |
| Terminal-Bench Hard | Artificial Analysis | 43.2% | 37.1% |
| τ²-Bench Telecom (AA) | Artificial Analysis | 98.3% | 80.7% |
| BFCL Overall | Berkeley Function Calling Leaderboard | – | 63.0% |
| τ²-bench | τ²-bench | 81.0% | – |
| τ³-bench Banking | τ²-bench | 9.8% | – |
| APEX-Agents (AA) | Artificial Analysis | 14.4% | – |
| Terminal-Bench 2.0 (Vals) | Vals AI | 49.4% | – |
| DeepResearch Bench | DeepResearch Bench | – | 45.2%† |
| METR Time Horizons | METR | – | 63.6% |
| Vending-Bench 2 | Andon Labs | 4432.1† | – |
| BrowseComp-Plus | BrowseComp-Plus | – | 50.5% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | – | 33.3%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 80.0%† | 83.9%† |
| MATH Level 5 | Epoch AI Benchmarking Hub | – | 97.8%† |
| AIME (Vals) | Vals AI | 91.7% | 85.3%† |
| MGSM | Vals AI | – | 91.8%† |
| AIME 2026 | MathArena | 96.7%† | – |
| HMMT February 2026 | MathArena | 86.4%† | – |
| USAMO 2026 | MathArena | 35.1%† | – |
| IMO 2025 | MathArena | – | 16.7%† |
| MathArena Apex | MathArena | 10.9%† | – |
| Omni-MATH (HELM) | HELM Capabilities | – | 71.4% |
| MATH 500 | Vals AI | – | 94.6%† |
| Arena Maths | Arena (LMArena) | 1444† | 1448 |
| AIME 2025 (AA) | Artificial Analysis | – | 88.3% |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | – | 49.4%† |
| MMLU-Pro | Vals AI | 86.0% | 85.6%† |
| LegalBench | Vals AI | 84.1% | 83.8%† |
| CorpFin | Vals AI | 62.9% | 59.7%† |
| TaxEval | Vals AI | 70.0% | 74.6%† |
| MedQA | Vals AI | 94.3% | 96.1%† |
| PRBench Finance | Scale AI SEAL | – | 47.7% |
| PRBench Legal | Scale AI SEAL | – | 48.6% |
| MultiNRC | Scale AI SEAL | – | 45.5%† |
| MMLU-Pro (HELM) | HELM Capabilities | – | 85.9% |
| CaseLaw v2 | Vals AI | 52.5% | – |
| MedCode | Vals AI | – | 47.3%† |
| MedScribe | Vals AI | – | 76.7%† |
| Instruction following | |||
| IFBench | Artificial Analysis | 72.3% | 71.4% |
| MultiChallenge | Scale AI SEAL | – | 56.6%† |
| TutorBench | Scale AI SEAL | – | 52.1%† |
| IFEval (HELM) | HELM Capabilities | – | 86.9% |
| WildBench (HELM) | HELM Capabilities | – | 86.1% |
| Arena Instruction Following | Arena (LMArena) | 1447† | 1403 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1457† | 1432 |
| Arena Creative Writing | Arena (LMArena) | 1448† | 1383 |
| Arena Multi-turn | Arena (LMArena) | 1471† | 1420 |
| Arena Longer Queries | Arena (LMArena) | 1468† | 1411 |
| Lech Mazur Creative Writing | Lech Mazur | – | 83.9%† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | – | 1214 |
| MMMU-Pro | Artificial Analysis | – | 70.1% |
| VISTA | Scale AI SEAL | – | 50.1%† |
| MMMU (validation) | MMMU | – | 82.9% |
| MMMU-Pro (official) | MMMU | – | 76.4% |
| MMMU-Pro (Vals) | Vals AI | – | 80.4%† |
| MortgageTax | Vals AI | – | 65.7%† |
| VTB | Scale AI SEAL | – | 13.7% |
| GeoBench | GeoBench | – | 60.0%† |
| VPCT | Epoch AI Benchmarking Hub | – | 52.0%† |
| Long context | |||
| Fiction.LiveBench 120k | Fiction.live | – | 100.0%† |
| AA-LCR v1.1 | Artificial Analysis | 75.7% | 74.7% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 145.8† | 146.8 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 27.9 | 20.2 |
| AA Progress Index v1 | Artificial Analysis | 280.9 | 258.3 |
| HELM Capabilities mean | HELM Capabilities | – | 81.1% |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 16.0% |
| MASK | Scale AI SEAL | – | 84.5%† |
| PropensityBench | Scale AI SEAL | – | 10.5% |
Data as of 2026-10-29. Best configuration of each model; every score links to its source on the model pages.
What changed
- GLM-5 (thinking): BenchLeader Index 60.6 → 60.1.
- GLM-5 (thinking): 1 new result (AA Progress Index v1); cached input price moved from $0.15/M to $0.20/M (34% up).
- o3: 1 new result (AA Progress Index v1).
- GLM-5 (thinking): cached input price moved from $0.20/M to $0.15/M (25% down).
- GLM-5 (thinking): cached input price moved from $0.15/M to $0.20/M (34% up).
- GLM-5 (thinking): cached input price moved from $0.20/M to $0.15/M (25% down).
- GLM-5 (thinking): cached input price moved from $0.15/M to $0.20/M (34% up).
- GLM-5 (thinking): cached input price moved from $0.20/M to $0.15/M (25% down).
- o3: 1 score revision.
- GLM-5 (thinking): 1 new result (τ³-bench Banking).
Related comparisons
- GLM-5 vs Claude Opus 5.5
- GLM-5 vs Gemini 4 Carbon
- GLM-5 vs GPT-6 Astra
- o3 vs Claude Opus 5.5
- o3 vs Gemini 4 Carbon
- o3 vs GPT-6 Astra
- o3 vs GPT-6.1 Sol
- GLM-5 vs Claude Fable 5.1
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GLM-5 vs o3: questions
- Is GLM-5 better than o3?
- GLM-5 (thinking) and o3 are neck and neck on quality (60.2 vs 61.0, within the ±4.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 29 Oct 2026, so check the category scores for your use.
- GLM-5 vs o3: which is cheaper, and will that change?
- GLM-5 is cheaper today: $1.55 against $3.50 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is GLM-5 or o3 better for coding?
- o3 scores higher in coding (62 vs 59 on the category index, where 50 is average).
- Which is faster, GLM-5 or o3?
- o3 streams ~2× faster (151 vs 69 output tokens per second, Artificial Analysis 7-day median).
- Is GLM-5 or o3 better for agents and tool use?
- o3 scores higher in agentic tasks (70 vs 58 on the category index, where 50 is average). On Terminal-Bench Hard, the top shared agents & tools benchmark, it is 43.2% for GLM-5 against 37.1% for o3.
More questions (4)
- Which is better for long documents?
- They are level on long-context benchmarks (63 vs 62). Both take about 205k tokens: GLM-5 204,800, o3 200,000.
- Which has lower latency?
- o3 answers first: 5.6 s against 46.9 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 vs o3?
- Across the 9 benchmarks both have, GLM-5 leads 8, o3 leads 1. Every overall index that rates both puts GLM-5 ahead.
- When was GLM-5 released?
- GLM-5 was released on 11 Feb 2026.