Claude Sonnet 5 vs GLM 5.3
Full profiles: Claude Sonnet 5 (Anthropic) and GLM 5.3 (Zhipu AI): every benchmark, price and rank.
Updated · Data from 11 sources
As of 6 Nov 2026, Claude Sonnet 5 and GLM 5.3 are neck and neck on quality (59.3 vs 59.0 on the BenchLeader Index, inside the ±3.7 margin), and GLM 5.3 is 1.9× cheaper at $2.15 vs $4.00 per 1M tokens blended. Choose Claude Sonnet 5 for reasoning (66 vs 55) and maths (65 vs 60) and fast first answers (10.8 s vs 101.7 s); choose GLM 5.3 for agents & tools (58 vs 55), human preference (67 vs 65), and long context (65 vs 63).
Key facts (as of 6 Nov 2026)
- Claude Sonnet 5 scores 59.3 on the BenchLeader Index and GLM 5.3 scores 59.0 (BenchLeader, 6 Nov 2026).
- Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 6 Nov 2026).
- GLM 5.3 costs $2.15 per 1M tokens blended (list price, 6 Nov 2026).
- Of the 21 benchmarks both models have, GLM 5.3 leads 19 (BenchLeader, 6 Nov 2026).
- On Arena WebDev, Claude Sonnet 5 is rated 1540 and GLM 5.3 1623 (LMArena, as published by 6 Nov 2026).
Claude Sonnet 5 benchmarks 
GLM 5.3 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 | Either | 60 vs 60 category score; Arena WebDev 1540 vs 1623 | Either |
| Agents & tool use | GLM 5.3 | 58 vs 55 category score; Arena Agent 1.1 vs 2.4 | Medium |
| Maths | Claude Sonnet 5 | 65 vs 60 category score; Arena Maths 1475 vs 1499 | Medium |
| Reasoning | Claude Sonnet 5 | 66 vs 55 category score; Arena Hard Prompts 1491 vs 1505 | High |
| Long documents | GLM 5.3 | 65 vs 63 category score; AA-LCR v1.1 79.7% vs 76.7% | Medium |
| Chat quality (human preference) | GLM 5.3 | 67 vs 65 on Arena votes | Medium |
| Lowest cost | GLM 5.3 | $2.15 vs $4.00 per 1M blended; agentic coding session $0.102 vs $0.160 | – |
| Fastest first answer | Claude Sonnet 5 | 10.8 s (Claude Sonnet 5 (high)) vs 101.7 s (GLM 5.3 (max)) 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
- Claude Sonnet 5 (high) and GLM 5.3 (max) are neck and neck on quality (59.3 vs 59.0, within the ±3.7 margin).
- Claude Sonnet 5 (high) is stronger in coding, maths, multimodal, reasoning.
- GLM 5.3 (max) is stronger in agents & tools, human preference, long context, knowledge.
- GLM 5.3 (max) is 1.9× cheaper ($2.15 vs $4.00 per 1M blended).
- Claude Sonnet 5 (high) streams 1.4× faster (86 vs 63 tokens per second).
At a glance
| Spec | Claude Sonnet 5 | GLM 5.3 |
|---|---|---|
| Developer | Anthropic | Zhipu AI |
| Released | 30 Jun 2026 | 18 Aug 2026 |
| Context window (tokens) | 1,000,000 | 1,000,000 |
| Compared at | Claude Sonnet 5 (high) | GLM 5.3 (max) |
| Open weights | No | Yes |
| Price in / out / cached input per 1M | $2.00 / $10.00 / $0.200 | $1.40 / $4.40 / $0.260 |
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 | Claude Sonnet 5 (high) | GLM 5.3 (max) | Leader |
|---|---|---|---|
| Reasoning | 65.5 | 55.3 | Claude Sonnet 5 |
| Coding | 60.4 | 60.0 | level |
| Agents & tools | 54.7 | 57.9 | GLM 5.3 |
| Maths | 65.4 | 60.2 | Claude Sonnet 5 |
| Knowledge | – | 53.9 | not comparable |
| Human preference | 64.8 | 67.1 | GLM 5.3 |
| Multimodal | 61.0 | – | not comparable |
| Long context | 63.1 | 64.6 | GLM 5.3 |
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: Knowledge, Instruction following, Multimodal.
ModelIndex
Where sources disagree
Across the 21 benchmarks both have, Claude Sonnet 5 leads 1, GLM 5.3 leads 19 and 1 is a tie. Every overall index that rates both puts GLM 5.3 ahead.
| Overall index | Claude Sonnet 5 (high) | GLM 5.3 (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 59.3 | 59.0 | neck and neck |
| Arena Text | 1462 | 1481 | GLM 5.3 |
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 | Claude Sonnet 5 (high) | GLM 5.3 (max) |
|---|---|---|
| Blended price per 1M tokens | $4.00 | $2.15 |
| Input / output per 1M | $2.00 / $10.00 | $1.40 / $4.40 |
| Cached input per 1M | $0.200 | $0.260 |
| Price per task: chat reply | $0.0038 | $0.0019 |
| Price per task: summarize a 30-page report | $0.030 | $0.019 |
| Price per task: code edit | $0.027 | $0.015 |
| Price per task: agentic coding session | $0.160 | $0.102 |
| Price per task: structured extraction | $0.0060 | $0.0037 |
| Agentic coding session, input mostly cached | $0.079 | $0.050 |
| Cost per benchmark task (AA Index) | – | $5.01 |
| Output tokens per task (AA Index) | – | 119k |
| Of which reasoning | – | 85k |
| Time per task (AA Index) | – | 31 min |
| Output speed | 86 tok/s | 63 tok/s |
| Time to first answer | 10.8 s | 101.7 s |
| Time to a complete answer | 53.2 s | 224.6 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Sonnet 5 (high) scores 59.3 and takes 11 s to its first answer token; Claude Sonnet 5 scores 55.9 in 1 s, Claude Sonnet 5 (xhigh) scores 55.6 in 31 s. GLM 5.3 (max) scores 59.0 and takes 102 s to its first answer token; GLM 5.3 (low) scores 47.6 in 12 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
133 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 | Claude Sonnet 5 (high) | GLM 5.3 (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 59.3 | 59.0 |
| Agents & tools score | BenchLeader | 54.7 | 57.9 |
| Coding score | BenchLeader | 60.4 | 60.0 |
| Human preference score | BenchLeader | 64.8 | 67.1 |
| Long context score | BenchLeader | 63.1 | 64.6 |
| Maths score | BenchLeader | 65.4 | 60.2 |
| Multimodal score | BenchLeader | 61.0 | – |
| Reasoning score | BenchLeader | 65.5 | 55.3 |
| Knowledge score | BenchLeader | – | 53.9 |
| Blended price $/M | BenchLeader | $4.00 | $2.15 |
| Output speed | BenchLeader | 86 tok/s | 63 tok/s |
| Time to first answer | BenchLeader | 10.8 s | 101.7 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,000,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.5%† | 90.9% |
| SimpleBench | SimpleBench | 60.6%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1491 | 1505 |
| LiveBench 1.0 Reasoning | LiveBench | 88.7%† | 85.8%† |
| GPQA Diamond (AA) | Artificial Analysis | 91.1%† | 91.7% |
| Humanity's Last Exam (AA) | Artificial Analysis | 35.7% | 42.3% |
| HLE-Diamond (AA) | Artificial Analysis | – | 17.9% |
| GPQA Diamond (Vals) | Vals AI | 88.9%† | 88.1% |
| CritPt | Artificial Analysis | 15.1% | 19.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 33.5% |
| MysteryMechanism | Vals AI | – | 23.0% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 34.0%† | 37.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 16.0%† | 33.0% |
| LMCA | Epoch AI Benchmarking Hub | 50.0% | 55.5%† |
| DTBench | Epoch AI Benchmarking Hub | 84.5% | 87.7%† |
| ForecastBench | ForecastBench | 63.5%† | – |
| BTF-3 | Epoch AI Benchmarking Hub | 14.2%† | 14.9%† |
| Coding | |||
| SciCode | SciCode | 54.3% | 59.0% |
| WeirdML | WeirdML | 68.8% | 75.4% |
| FrontierCode | Cognition | 42.7%† | 40.1% |
| GSO-Bench | GSO-Bench | 37.3%† | – |
| Arena Coding | Arena (LMArena) | 1520 | 1520 |
| Arena WebDev | Arena (LMArena) | 1540 | 1623 |
| LiveBench 1.0 Coding | LiveBench | 80.7%† | 79.0%† |
| LiveCodeBench | Vals AI | 82.4%† | 80.5% |
Show all 133 benchmarks
| Metric | Publisher | Claude Sonnet 5 (high) | GLM 5.3 (max) |
|---|---|---|---|
| Coding | |||
| IOI | Vals AI | 45.0%† | 68.4% |
| SWE-bench (Vals) | Vals AI | 79.6%† | 95.4% |
| Code Migration | Vals AI | 44.4%† | 44.2% |
| Code Migration v1.1 | Vals AI | – | 55.1% |
| ProgramBench | Vals AI | 0.0%† | 1.5% |
| Vibe Code Bench 1-100 | Vals AI | 13.8%† | 20.0% |
| Vibe Code Bench v1.1 | Vals AI | 81.3%† | 78.1% |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 60.0%† | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 48.2% | 69.0% |
| CursorBench | Cursor | 30.8% | 42.6% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1463.1 | 1317.4† |
| FrontierSWE | Epoch AI Benchmarking Hub | – | 30.2% |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 5.0% | 41.9% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 80.5%† | 83.9% |
| Startrise Frontend | Startrise | 66.6%† | 73.2%† |
| GameBench 2 Overall | DuelLab GameBench 2 | 46.4† | 64.1† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 54.4† | 72.1 |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 40.1% |
| FrontierCode 2.0 (AA) | Artificial Analysis | 17.8%† | 21.6% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 12.4%† | 41.8% |
| APEX-Agents | Mercor | 54.5%† | 56.6%† |
| Arena Agent | Arena (LMArena) | 2.4 | 1.1 |
| LiveBench Agentic Coding | LiveBench | 59.4%† | 60.9%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | 74.5%† | 71.5% |
| MCP Atlas | Scale AI SEAL | – | 84.2%† |
| HiL-Bench | Scale AI SEAL | – | 58.3%† |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 57.5% |
| GDPval-AA v2.1 | Artificial Analysis | – | 1657 |
| τ³-Banking (AA) | Artificial Analysis | 37.3%† | 50.3% |
| ITBench SRE (AA) | Artificial Analysis | – | 46.1% |
| Analyst Agent (AA) | Artificial Analysis | 46.3%† | 38.8% |
| CyberBench | Vals AI | 61.9%† | 72.1% |
| Finance Agent v2 | Vals AI | 53.9%† | 55.8% |
| Finance Agent v2.1 | Vals AI | – | 65.0% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 14.6% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 5.0%† | 8.3% |
| Legal Research Bench | Vals AI | 41.8%† | 49.0% |
| SkillsBench | Vals AI | 46.5%† | 47.5% |
| SREBench | Vals AI | 0.4%† | 9.2% |
| Tax Agent Bench | Vals AI | 29.4%† | 39.7% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 10.6%† | 38.9% |
| Terminal-Bench Science | Vals AI | 5.7%† | 5.7% |
| GBAEval | Epoch AI Benchmarking Hub | 65.3%† | – |
| Vending-Bench 2 | Andon Labs | 6377.7† | 8163.6† |
| AutomationBench | Artificial Analysis | 32.1% | 62.2% |
| GDP.pdf (AA) | Artificial Analysis | 9.4% | 11.2% |
| Harvey LAB-AA v1.1 | Harvey | – | 0.3% |
| EnterpriseOps-Gym | Artificial Analysis | 43.9%† | 36.4% |
| AA-Briefcase v1.1 | Artificial Analysis | 1356† | 1503 |
| GDPx | GDPx by SemiAnalysis | – | 2.7% |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 9.5% |
| CWE-Bench (AA) | Artificial Analysis | – | 55.8% |
| DeepSec-Bench (AA) | Artificial Analysis | – | 24.3% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 29.0% |
| FACTS Search | – | 87.7%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 65.6%† | 68.8% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 29.3%† | 29.3% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 94.7%† | 91.1% |
| ProofBench | Vals AI | 77.0%† | 49.0% |
| LiveBench Mathematics | LiveBench | 92.9%† | 87.9%† |
| Arena Maths | Arena (LMArena) | 1475 | 1499 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 32.9%† | 41.0% |
| LiveBench 1.0 Data Analysis | LiveBench | 71.7%† | 70.2%† |
| AA-Omniscience v1.1 | Artificial Analysis | – | 6.0 |
| MMLU-Pro | Vals AI | 87.5%† | 86.8% |
| LegalBench | Vals AI | 83.9%† | 84.8% |
| CorpFin | Vals AI | 67.0%† | – |
| TaxEval | Vals AI | 75.6%† | 72.4% |
| Excel Modeling Benchmark | Vals AI | 66.3%† | 56.3% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 69.1% |
| MedCode | Vals AI | 47.5%† | 42.9% |
| MedScribe | Vals AI | 76.0%† | 88.8% |
| Public Benefits Bench v1.1 | Vals AI | 66.0%† | 68.5% |
| DrugDiscoveryBench | Scale AI SEAL | 50.0%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 32.3% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 61.1% |
| HealthBench Professional (AA) | OpenAI | – | 52.8% |
| Instruction following | |||
| LiveBench Language | LiveBench | 75.0%† | 79.9%† |
| LiveBench Instruction Following | LiveBench | 63.9%† | 69.3%† |
| Arena Instruction Following | Arena (LMArena) | 1467 | 1479 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1462 | 1481 |
| EQ-Bench 4 | EQ-Bench | 1221† | – |
| Arena Creative Writing | Arena (LMArena) | 1436 | 1457 |
| Arena Multi-turn | Arena (LMArena) | 1472 | 1478 |
| Arena Longer Queries | Arena (LMArena) | 1481 | 1486 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1273 | – |
| MMMU-Pro | Artificial Analysis | 77.3%† | – |
| MMMU-Pro (Vals) | Vals AI | 83.0%† | – |
| MortgageTax | Vals AI | 70.0%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 24.9%† | – |
| Code2Video-AA | Artificial Analysis | – | 1034 |
| AA-3D | Artificial Analysis | – | 783 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 76.7% | 79.7% |
| Arena Document | Arena (LMArena) | 1476 | – |
| MLCR | Artificial Analysis | 4.4% | 17.2% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 156.7† | 156.9† |
| LiveBench 1.0 | LiveBench | 76.0%† | 76.1%† |
| AA Intelligence Index v5.0 | Artificial Analysis | – | 31.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 31.7 | 44.8 |
| AA Progress Index v1 | Artificial Analysis | 296.8 | 322.7 |
| Vals Index | Vals AI | – | 59.7 |
| Vals Finance Index | Vals AI | – | 60.5% |
| Vals RSI Index | Vals AI | – | 22.9% |
| Vals Multimodal Index | Vals AI | 68.8%† | – |
| Vals Router Index | Vals AI | – | 59.7%† |
| AA Cyber Index | Artificial Analysis | – | 36.4 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 28.2% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 33.3 |
Data as of 2026-11-06. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261103-09b2218cbd85 (https://benchmarks.duellab.org/releases/gb2-20261103-09b2218cbd85/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- GLM 5.3 (max): 1 score revision; cached input price moved from $0.30/M to $0.26/M (13% down).
- GLM 5.3 (max): cached input price moved from $0.26/M to $0.30/M (15% up).
- GLM 5.3 (max): cached input price moved from $0.30/M to $0.26/M (13% down).
- GLM 5.3 (max): cached input price moved from $0.26/M to $0.30/M (15% up).
- Claude Sonnet 5 (high): 1 score revision.
- GLM 5.3 (max): BenchLeader Index 61.3 → 58.9.
- Claude Sonnet 5 (high): 1 new result (AA Progress Index v1).
- GLM 5.3 (max): 5 new results (AA Intelligence Index v5.0, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $2.01/M to $5.01/M (150% up); cached input price moved from $0.30/M to $0.26/M (13% down).
- GLM 5.3 (max): cached input price moved from $0.26/M to $0.30/M (15% up).
- GLM 5.3 (max): cached input price moved from $0.30/M to $0.26/M (13% down).
Related comparisons
- Claude Sonnet 5 vs Claude Opus 5.5
- Claude Sonnet 5 vs Gemini 4 Carbon
- Claude Sonnet 5 vs GPT-6 Astra
- GLM 5.3 vs Claude Opus 5.5
- GLM 5.3 vs Gemini 4 Carbon
- GLM 5.3 vs GPT-6 Astra
- GLM 5.3 vs GPT-6.1 Sol
- Claude Sonnet 5 vs Claude Fable 5.1
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Sonnet 5 vs GLM 5.3: questions
- Is Claude Sonnet 5 better than GLM 5.3?
- Claude Sonnet 5 (high) and GLM 5.3 (max) are neck and neck on quality (59.3 vs 59.0, within the ±3.7 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 6 Nov 2026, so check the category scores for your use.
- Claude Sonnet 5 vs GLM 5.3: which is cheaper, and will that change?
- GLM 5.3 is cheaper today: $2.15 against $4.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Sonnet 5 or GLM 5.3 better for coding?
- They are level in coding (60 vs 60 on the category index). On Arena WebDev, the top shared coding benchmark, it is 1540 for Claude Sonnet 5 against 1623 for GLM 5.3.
- Which is faster, Claude Sonnet 5 or GLM 5.3?
- Claude Sonnet 5 streams faster (86 vs 63 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Sonnet 5 or GLM 5.3 better for agents and tool use?
- GLM 5.3 scores higher in agentic tasks (58 vs 55 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is 2.4 for Claude Sonnet 5 against 1.1 for GLM 5.3.
More questions (4)
- Which is better for long documents?
- GLM 5.3 scores higher on long-context benchmarks (65 vs 63). Both accept 1,000,000 tokens of context.
- Which has lower latency?
- Claude Sonnet 5 answers first: 10.8 s against 101.7 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 Claude Sonnet 5 vs GLM 5.3?
- Across the 21 benchmarks both have, Claude Sonnet 5 leads 1, GLM 5.3 leads 19 and 1 is a tie. Every overall index that rates both puts GLM 5.3 ahead.
- When was GLM 5.3 released?
- GLM 5.3 was released on 18 Aug 2026.