Compare / head-to-head
Claude Opus 4.7vsGLM-5.3
Claude Opus 4.7 leads 5 of 6 shared benchmarks. GLM-5.3 is 5.2x cheaper per token. Both offer a 1M-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
5 – 1
Claude Opus 4.7 leads
Cheaper per token
GLM-5.3
5.2x cheaper, input + output
Larger context
Tie
both 1M tokens
Provider uptime (30d)
100% · —
Anthropic · Z.ai
Claude Opus 4.7
Anthropic · released 2026-04-16
textvision
Quality
Benchmark matrix
6 shared · 11 only Claude Opus 4.7 · 18 only GLM-5.3| Benchmark | Claude Opus 4.7 | GLM-5.3 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 6 | ||||
| GPQA Diamond | 94.2% ↗ | 90.9% ↗epoch | +3.3 pt | Claude Opus 4.7 |
| FrontierMath Tier 4 v2 (Epoch AI run) | 31.7% ↗epoch | 29.3% ↗epoch | +2.4 pt | Claude Opus 4.7 |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 70.2% ↗epoch | 68.8% ↗epoch | +1.4 pt | Claude Opus 4.7 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 97.8% ↗epoch | 91.1% ↗epoch | +6.7 pt | Claude Opus 4.7 |
| SimpleQA Verified | 51.7% ↗epoch | 41% ↗epoch | +10.7 pt | Claude Opus 4.7 |
| Terminal-Bench 2.1 | 66.1% ↗ | 88.2% ↗ | -22.1 pt | GLM-5.3 |
| Only Claude Opus 4.7 reports · 11 | ||||
| SWE-bench Verified | 87.6% ↗ | not reported | — | — |
| AIME 2026 | 95.8% ↗matharena ⚠ | not reported | — | — |
| Humanity's Last Exam (no tools) | 46.9% ↗ | not reported | — | — |
| LMArena Elo | 1483.4 ↗ | not reported | — | — |
| BrowseComp | 79.8% ↗ | not reported | — | — |
| Humanity's Last Exam (with tools) | 54.7% ↗ | not reported | — | — |
| OSWorld-Verified | 82.8% ↗ | not reported | — | — |
| SWE-bench Multilingual | 80.5% ↗ | not reported | — | — |
| SWE-bench Multimodal | 34.5% ↗ | not reported | — | — |
| SWE-Bench Pro | 64.3% ↗ | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 83.5% ↗epoch | not reported | — | — |
| Only GLM-5.3 reports · 18 | ||||
| Agents' Last Exam | not reported | 28.5% ↗ | — | — |
| Agents' Last Exam (ALE-CLI) | not reported | 28.5% ↗ | — | — |
| AutomationBench v1.0.6 | not reported | 48.2% ↗ | — | — |
| CyberGym | not reported | 84.5% ↗ | — | — |
| DeepSWE v1.1 | not reported | 66.9% ↗ | — | — |
| ExploitBench | not reported | 54.4% ↗ | — | — |
| ExploitGym 2h | not reported | 105 ↗ | — | — |
| ExploitGym 6h | not reported | 130 ↗ | — | — |
| FrontierSWE | not reported | 78.1% ↗ | — | — |
| GDPval-AA v2 | not reported | 1769 ↗ | — | — |
| HLE (with tools) | not reported | 62.5% ↗ | — | — |
| NL2Repo | not reported | 58% ↗ | — | — |
| PostTrainBench | not reported | 39.8% ↗ | — | — |
| ProgramBench (Almost Solved) | not reported | 19% ↗ | — | — |
| SWE-Marathon v1.1 | not reported | 42.5% ↗ | — | — |
| Terminal-Bench 3.0 | not reported | 28.3% ↗ | — | — |
| Toolathlon Verified | not reported | 73% ↗ | — | — |
| Z.ai Code Bench (max effort) | not reported | 34.5% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Claude Opus 4.7 minus GLM-5.3 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Claude Opus 4.7 | GLM-5.3 | Edge |
|---|---|---|---|
| Context window | 1M tokens | 1M tokens | Tie |
| Max output | 128K tokens | 131K tokens | GLM-5.3 |
| Input price / 1M | $5 ↗ | $1.4 ↗ | GLM-5.3 |
| Output price / 1M | $25 ↗ | $4.4 ↗ | GLM-5.3 |
| Cached input / 1M | $0.5 ↗ | $0.26 ↗ | GLM-5.3 |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $30 | $5.8 | GLM-5.3 |
| Modalities | text · vision | text | Claude Opus 4.7 |
| Released | 2026-04-16 | 2026-08-14 | — |
| Cited benchmark scores | 18 | 33 | — |
Reliability
Provider status
Last 60 days · refreshed daily on this page · live on /statusMore matchups
Claude Opus 4.7 vs …
Models sharing the most benchmarksMore matchups
GLM-5.3 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Opus 4.7 and GLM-5.3 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.