Model comparison
GLM 5.3 vs Grok 4.7
Z.ai against SpaceXAI, compared on context, price, and verified benchmark results.
Catalog record checked September 26, 2026Individual provider fields may changeEvidence confidence: High — see receipts below
Z.ai
GLM 5.3
Frontier · Open weights
SpaceXAI
Grok 4.7
Frontier
| Specification | GLM 5.3 | Grok 4.7 |
|---|---|---|
| Provider | ||
| Provider | Z.ai | SpaceXAI |
| Tier | ||
| Tier | Frontier | Frontier |
| Context window | ||
| Context window | Winner: 1M | 500K |
| Max output | ||
| Max output | 128K | Not verifiedUnverified |
| Input / 1M tokens | ||
| Input / 1M tokens | Winner: $1.40 | $2 |
| Output / 1M tokens | ||
| Output / 1M tokens | Winner: $4.40 | $6 |
| Weights | ||
| Weights | Open | Closed |
| Parameters | ||
| Parameters | 753B total (MoE; active count unpublished) | new, larger base model with longer RL on multi-hour tasks |
| Reasoning levels | ||
| Reasoning levels | low, high, max | low, medium, high, xhigh |
| Modalities | ||
| Modalities | text | text, image |
| License | ||
| License | glm-5.3 (custom) | Not disclosedUnverified |
| API model id | ||
| API model id | glm-5.3 | grok-4.7 |
| Released | ||
| Released | August 14, 2026 | September 21, 2026 |
| Artificial Analysis Intelligence Index [max] (2026-09-26) | ||
| Artificial Analysis Intelligence Index [max] (2026-09-26) | 44.8 | Not verifiedUnverified |
| Artificial Analysis Intelligence Index [xhigh] (2026-09-26) | ||
| Artificial Analysis Intelligence Index [xhigh] (2026-09-26) | Not verifiedUnverified | 46.4 |
| Terminal-Bench 2.1 (2026-08-14) | ||
| Terminal-Bench 2.1 (2026-08-14) | 88.2 | Not verifiedUnverified |
| DeepSWE 1.1 (2026-08-14) | ||
| DeepSWE 1.1 (2026-08-14) | 66.9 | Not verifiedUnverified |
| DeepSWE 1.1 [high] (2026-09-21) | ||
| DeepSWE 1.1 [high] (2026-09-21) | Not verifiedUnverified | 71 |
| Humanity's Last Exam (2026-08-14) | ||
| Humanity's Last Exam (2026-08-14) | 62.5 | Not verifiedUnverified |
| SWE-bench Verified (2026-09-01) | ||
| SWE-bench Verified (2026-09-01) | 95.4 | Not verifiedUnverified |
| Terminal-Bench 4.0 [xhigh] (2026-09-26) | ||
| Terminal-Bench 4.0 [xhigh] (2026-09-26) | Not verifiedUnverified | 25.8 |
Prices are USD per million tokens at standard rates, excluding batch and caching discounts. Bold indicates the better figure where one is objectively better. Values we could not confirm from the provider are shown as “Not verified” rather than estimated.
Benchmark receipts
Where each score comes from, and how far it can be compared across models.
- 2026-09-26Artificial Analysis Terminal-Bench 4.0 (independent AA run, part of Intelligence Index v4.3.2)
Agentic terminal work: long multi-step tasks executed in a sandboxed shell environment.
Comparable with caveatNOT comparable with Terminal-Bench 2.x or 3.0 — 4.0 uses a new, non-overlapping task set. Within 4.0, scores from different harnesses (tbench.ai agent entries vs Artificial Analysis runs) are not interchangeable.
- 2026-09-26Artificial Analysis
Composite index blending reasoning, knowledge, and coding evals into one 0–100 score.
Comparable with caveatOnly same-version scores are comparable. v4.3.2 was rebaselined (the top score fell from 66 on v4.1.1 to ~58), so v4.1.1 figures must not be compared with v4.3.2 figures — check measuredAt.
- 2026-09-21xAI — Grok 4.7 announcement (vendor; not yet on Datacurve's board)
Agentic software engineering: pass@1 on real repo tasks with reported cost per completed task.
Directly comparable
- 2026-09-01vals.ai SWE-bench Verified leaderboard (independent, mini-swe-agent bash-only); Zhipu docs list 77.8 inherited from GLM-5 — attribution contested, both noted
Real GitHub issue resolution: does the model's patch pass the hidden tests.
Comparable with caveatPost-audit vendor claims and pre-audit scores sit on different task trust levels; scaffolding (agent harness, compute budget) also dominates results. Never aggregate across scaffolds.
- 2026-08-14Z.ai GLM-5.3 blog + HF model card (vendor, with tools, full set)
Agentic terminal work: multi-step tasks executed in a sandboxed shell environment.
Comparable with caveatNot comparable with Terminal-Bench 3.0 or 4.0 (different task sets) or v1; harness configuration (container, time limits) also shifts results.
Pricing tiers
GLM 5.3: Token API listed at $1.40/$4.40 per MTok from 2026-08-18 (cached input $0.26); Coding Plan access on a points quota. 1M context, 128K max output per official docs; thinking always on. 753B total MoE per the Hugging Face card (active count unpublished).
Grok 4.7: $2/$6 per MTok (<200K prompt); $4/$12 for all tokens at ≥200K. Cached input $0.50/MTok ($1.00 at ≥200K). US regional endpoint 1.1x. No Batch API; the 2x-price 'Fast' variant is Cursor and Grok Build only. xAI lists no separate text-output cap. Knowledge cutoff May 2026.
GLM 5.3
GLM 5.3 is Zhipu's flagship (~753B MoE), near the top of the leaderboards, and the current GLM Coding Plan default.
Best for
- Coding Plan subscribers
- Long-horizon coding
- Chinese + English
Watch out
Open weights dropped 2026-08-28 under Z.ai's custom glm-5.3 licence (not a standard open-source licence — review before commercial use; secondary coverage says >$10B-revenue providers need a security review). 5.2/5.1 Coding Plan requests route to 5.3.
Grok 4.7
Grok 4.7 is xAI's flagship since 2026-09-21 — a larger new base model with longer RL on multi-hour tasks, at Grok 4.6's $2/$6 price.
Best for
- Coding agents
- Cost-sensitive frontier work
- Cursor and GitHub Copilot users
Watch out
Token rates double above a 200K-token prompt and there is no Batch API; the DeepSWE figure is vendor-reported.
When the cheaper one wins
GLM 5.3 is cheaper on output at $4.40 per million tokens against $6 for Grok 4.7 — about 1.4×. Use the cheaper tier for classification, extraction, summarisation, and any task where the expensive model’s extra score does not change the accepted output. The expensive one only pays if your hardest task actually fails on the cheap tier. These are standard-tier API rates, excluding batch and cache discounts.
Run the model pickerEvidence confidence: High
How strong and complete the evidence behind this comparison is — not a prediction of which model is better.
- Pricing verified on both sides — Input and output rates are verified for both models.
- 4/5 core specs verified on both sides — Not published for at least one side: max output.
- 2 shared named benchmarks with differing scores — Measured on: Artificial Analysis Intelligence Index, DeepSWE 1.1.
- Verified within the last 90 days — Newest catalog check was 2 days ago.
- Both models carry source citations — Each side has at least two catalog sources on record.
Source receipts
Each catalog figure was checked against the provider or an independent second source on the date shown.
- GLM 5.3: Z.ai GLM-5.3 announcement (accessed 2026-08-29)
- GLM 5.3: Hugging Face — zai-org/GLM-5.3 (weights, 2026-08-28) (accessed 2026-08-30)
- Grok 4.7: xAI — Grok 4.7 (accessed 2026-09-26)
- Grok 4.7: xAI docs — grok-4.7 model card (accessed 2026-09-26)
- Grok 4.7: xAI pricing (grok-4.7 $2/$6, ≥200K $4/$12) (accessed 2026-09-26)
- Grok 4.7: GitHub changelog — Grok 4.7 in Copilot (2026-09-21) (accessed 2026-09-26)
Related comparisons
Common questions
GLM 5.3 vs Grok 4.7
Answered from the verified figures on this page rather than general guidance.
Is GLM 5.3 or Grok 4.7 cheaper for input?
Is GLM 5.3 or Grok 4.7 cheaper for output?
Which has the larger context window, GLM 5.3 or Grok 4.7?
Do GLM 5.3 and Grok 4.7 support the same reasoning levels?
Should I use GLM 5.3 or Grok 4.7?
Can I self-host GLM 5.3 or Grok 4.7?
Next step
Choosing between them
The verified figures that separate this pair, computed from the catalog rather than restated boilerplate.
- Input price: GLM 5.3 costs $1.40 per 1M tokens versus $2 for Grok 4.7 — a 1.4x difference at the headline tier.
- Context: GLM 5.3 takes 1M against 500K for Grok 4.7 — only decisive if your prompts approach the smaller window.
- Measured capability: Grok 4.7 leads Artificial Analysis Intelligence Index 46.4 to 44.8 (measured 2026-09-26).
- Deployment: GLM 5.3 publishes weights you can self-host; the other is API-only.
Choosing a harness rather than a model — Cursor, Copilot, Claude Code, Windsurf, or Lovable? Compare agentic harnesses · Latest releases.