GLM-5.2 vs Grok 4.7 vs Qwen3.8 Max 0902
Too close to call on our weighted score (GLM-5.2 65, Qwen3.8 Max 0902 64, Grok 4.7 63). The right pick depends on what you value most.Too close to call on our weighted score (Qwen3.8 Max 0902 80, Grok 4.7 79, GLM-5.2 78). The right pick depends on what you value most.
Z.ai (Zhipu)
GLM-5.2
65/100- ECI153.1
- Price$1.40 / $4.40
- Context1M
xAI
Grok 4.7
63/100- ECI156.4
- Price$2.00 / $6.00
- Context500K
Alibaba (Qwen)
Qwen3.8 Max 0902
64/100- ECI156.0
- Price$2.00 / $6.00
- Context1M
Z.ai (Zhipu)
GLM-5.2
78/100- ECI153.1
- Price$1.40 / $4.40
- Context1M
xAI
Grok 4.7
79/100- ECI156.4
- Price$2.00 / $6.00
- Context500K
Alibaba (Qwen)
Qwen3.8 Max 0902
80/100- ECI156.0
- Price$2.00 / $6.00
- Context1M
Too close to call
It is close. Our weighted score puts them within a point (GLM-5.2 65/100, Qwen3.8 Max 0902 64/100, Grok 4.7 63/100), so choose by what matters most for your work: Grok 4.7 for raw capability and GLM-5.2 on price. The score weighs capability 59%, price 29%, context window 12%.It is close. Our weighted score puts them within 1 points (Qwen3.8 Max 0902 80/100, Grok 4.7 79/100, GLM-5.2 78/100), so choose by what matters most for your work: Grok 4.7 for raw capability and GLM-5.2 on price. The score weighs capability 83%, context window 17%.
- CapabilityGrok 4.7Capabilities Index (ECI): Grok 4.7 156.4 · Qwen3.8 Max 0902 156.0 · GLM-5.2 153.1
- Lowest priceGLM-5.2GLM-5.2 $2.15 · Grok 4.7 $3.00 · Qwen3.8 Max 0902 $3.00 per 1M tokens (3:1 blend)
- Longest contextGLM-5.2 and Qwen3.8 Max 0902GLM-5.2 1,000,000 · Qwen3.8 Max 0902 1,000,000 · Grok 4.7 500,000 tokens
- Widest inputsQwen3.8 Max 0902GLM-5.2: Text · Grok 4.7: Text, Images, PDFs · Qwen3.8 Max 0902: Text, Images, PDFs, Video
- Self-hostingGLM-5.2Publishes downloadable weights
| Measure | Weight | GLM-5.2 | Grok 4.7 | Qwen3.8 Max 0902 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 59% | 81 | 84 | 84 |
| Price | 29% | 33 | 26 | 26 |
| Context window | 12% | 64 | 53 | 64 |
| Overall | 100% | 65/100 | 63/100 | 64/100 |
Left out because at least one model lacks the data: inputs & features. The remaining weights were rescaled.
| Measure | Weight | GLM-5.2 | Grok 4.7 | Qwen3.8 Max 0902 |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 83% | 81 | 84 | 84 |
| Context window | 17% | 64 | 53 | 64 |
| Overall | 100% | 78/100 | 79/100 | 80/100 |
Left out because at least one model lacks the data: inputs & features. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Unavailable specifications are marked “Not published”. Publisher results include their evaluation settings and source.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 153.1 | 156.4 (best) | 156.0 |
| ECI rank | #55 of 159 | #39 of 159 (best) | #42 of 159 |
| ECI confidence interval | 151.3–154.9 | 154.5–158.5 | 153.9–158.3 |
| ECI evaluation date | Jun 16, 2026 | Sep 21, 2026 | Sep 1, 2026 |
| Documented capabilities | Open flagship GLM for long-horizon coding agents and million-token context work | xAI's frontier model for long-running agents, coding, knowledge work, and visual projects | 2026-09-02 upgraded snapshot of Qwen3.8 Max with stronger coding, collaborative agents, and multimodal document understanding |
| Model parameters | Not published | Not published | Not published |
| Output modalities | Text | Text | Text |
| GPQA DiamondGraduate-level science questions | 91.9%max | 92.7%xhigh (best) | 92.3%xhigh |
| FrontierMath Tiers 1–3Research-level mathematics | 59.2%max | 53.0%xhigh | 65.6%xhigh (best) |
| OTIS Mock AIME 2024–2025Competition mathematics | 86.4%max | 98.1%xhigh | 100%xhigh (best) |
| SWE-bench VerifiedFixing real GitHub issues | 78.7%max | Not published | Not published |
| SimpleQA VerifiedShort factual questions | 34.2%max | 56.0%xhigh (best) | 47.3%xhigh |
| Publisher-reported benchmarks | |||
| SWE-Bench ProPublisher-reported result | 62.1resolve rateReported resultJun 16, 2026 | Not published | Not published |
| Terminal-Bench 2.1Publisher-reported result | 82.7success rateReported resultJun 16, 2026 | Not published | Not published |
| FrontierSWEPublisher-reported result | 74.4dominanceReported resultJun 16, 2026 | Not published | Not published |
| Humanity's Last Exam · text-only subsetPublisher-reported result | 40.5accuracyReported resultJun 16, 2026 | Not published | Not published |
| Humanity's Last Exam · text-only subsetwith tools | 54.7accuracyReported resultJun 16, 2026 | Not published | Not published |
| CritPtPublisher-reported result | 20.9accuracyReported resultJun 16, 2026 | Not published | Not published |
| AIME 2026Publisher-reported result | 99.2accuracyReported resultJun 16, 2026 | Not published | Not published |
| HMMT November 2025Publisher-reported result | 94.4accuracyReported resultJun 16, 2026 | Not published | Not published |
| HMMT February 2026Publisher-reported result | 92.5accuracyReported resultJun 16, 2026 | Not published | Not published |
| IMOAnswerBenchPublisher-reported result | 91accuracyReported resultJun 16, 2026 | Not published | Not published |
| GPQA DiamondPublisher-reported result | 91.2accuracyReported resultJun 16, 2026 | Not published | Not published |
| NL2RepoPublisher-reported result | 48.9resolve rateReported resultJun 16, 2026 | Not published | Not published |
| DeepSWEPublisher-reported result | 46.2resolve rateReported resultJun 16, 2026 | Not published | Not published |
| Program BenchPublisher-reported result | 63.7scoreReported resultJun 16, 2026 | Not published | Not published |
| PostTrainBenchPublisher-reported result | 34.3scoreReported resultJun 16, 2026 | Not published | Not published |
| SWE MarathonPublisher-reported result | 13resolve rateReported resultJun 16, 2026 | Not published | Not published |
| MCP Atlas · public subsetPublisher-reported result | 76.8scoreReported resultJun 16, 2026 | Not published | Not published |
| Tool-DecathlonPublisher-reported result | 48.2scoreReported resultJun 16, 2026 | Not published | Not published |
| CursorBench 4.0xhigh effort | Not published | 46.3scoreReported resultSep 21, 2026 | Not published |
| DeepSWE 1.1high effort | Not published | 71scoreReported resultSep 21, 2026 | Not published |
| EEBenchxhigh effort | Not published | 64scoreReported resultSep 21, 2026 | Not published |
| AA-Briefcase 1.1xhigh effort | Not published | 1657EloReported resultSep 21, 2026 | Not published |
| Terminal-Bench 4.0xhigh effort | Not published | 37.6scoreReported resultSep 21, 2026 | Not published |
| Harvey Legal Agent Benchmarkxhigh effort | Not published | 19.6scoreReported resultSep 21, 2026 | Not published |
| HealthBenchxhigh effort; professional | Not published | 56.7scoreReported resultSep 21, 2026 | Not published |
| GDPvalxhigh effort | Not published | 1695EloReported resultSep 21, 2026 | Not published |
| Price per million tokens | |||
| Input | $1.40 (best) | $2.00 | $2.00 |
| Output | $4.40 (best) | $6.00 | $6.00 |
| Cached input | $0.26 (best) | $0.50 | Not published |
| Blended (3:1) | $2.15 (best) | $3.00 | $3.00 |
| Long-context rate | Same rate | Over 200K: $4.00 / $12.00 | Same rate |
| Price source | Official Z.AI API | Official xAI API | Median of 9 providers |
| Limits | |||
| Context window | 1,000,000 tokens (best) | 500,000 tokens | 1,000,000 tokens (best) |
| Max input | Not published | Not published | Not published |
| Max output | 131,072 tokens | 500,000 tokens (best) | 131,072 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | Yes |
| PDFs | No | Yes | Yes |
| Audio | No | No | No |
| Video | No | No | Yes |
| Reasoning | Yeshigh · max | Yeslow · medium · high · xhigh | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | Not published |
| File attachments | No | Yes | Yes |
| Temperature control | Yes | Yes | Yes |
| Reasoning control | effort | effort | always |
| Availability | |||
| Weights | Open | Proprietary | Proprietary |
| API model ID | glm-5.2 | grok-4.7 | Not published |
| API providers | 88 (best) | 35 | 9 |
| Released | Jun 13, 2026 | Sep 21, 2026 | Sep 2, 2026 |
| Knowledge cutoff | Not published | May 2026 | Not published |
| Publisher documentation | Publisher documentation · Model weights | Publisher documentation | Not published |
| Source metadata updated | Jun 13, 2026 | Sep 21, 2026 | Sep 2, 2026 |
| Data sources | models.dev · Epoch AI | models.dev · Epoch AI | models.dev · Epoch AI |
| Catalog checked | Nov 22, 2026 | Nov 22, 2026 | Nov 22, 2026 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
GLM-5.2$22.80
Grok 4.7$32.00
Qwen3.8 Max 0902$32.00
Which should you choose?
Which is better: GLM-5.2, Grok 4.7 or Qwen3.8 Max 0902?
It is close. Our weighted score puts them within a point (GLM-5.2 65/100, Qwen3.8 Max 0902 64/100, Grok 4.7 63/100), so choose by what matters most for your work: Grok 4.7 for raw capability and GLM-5.2 on price. The score weighs capability 59%, price 29%, context window 12%.It is close. Our weighted score puts them within 1 points (Qwen3.8 Max 0902 80/100, Grok 4.7 79/100, GLM-5.2 78/100), so choose by what matters most for your work: Grok 4.7 for raw capability and GLM-5.2 on price. The score weighs capability 83%, context window 17%.
Which is cheaper, GLM-5.2, Grok 4.7 or Qwen3.8 Max 0902?
GLM-5.2 is cheaper at $1.40 input / $4.40 output per million tokens (official Z.AI API price). Grok 4.7 costs $2.00 input / $6.00 output per million tokens (official xAI API price); Qwen3.8 Max 0902 costs $2.00 input / $6.00 output per million tokens (median across 9 API providers). At a typical mix of three input tokens to one output token, that is $2.15 per million tokens for GLM-5.2 versus $3.00 for Grok 4.7 (1.4× as much) and $3.00 for Qwen3.8 Max 0902 (1.4× as much).
Which scores higher on benchmarks?
Grok 4.7 scores higher on the Capabilities Index (ECI): Grok 4.7 156.4 (#39 of 159), Qwen3.8 Max 0902 156.0 (#42 of 159) and GLM-5.2 153.1 (#55 of 159). The confidence ranges of the top two overlap (154.5–158.5 vs 153.9–158.3), so treat the gap as small. On individual benchmarks: GPQA Diamond — Grok 4.7 92.7%, Qwen3.8 Max 0902 92.3%, GLM-5.2 91.9%; FrontierMath Tiers 1–3 — Qwen3.8 Max 0902 65.6%, GLM-5.2 59.2%, Grok 4.7 53.0%; OTIS Mock AIME 2024–2025 — Qwen3.8 Max 0902 100%, Grok 4.7 98.1%, GLM-5.2 86.4%; SimpleQA Verified — Grok 4.7 56.0%, Qwen3.8 Max 0902 47.3%, GLM-5.2 34.2%.
Which is better for coding?
There are no published SWE-bench Verified results for Grok 4.7 and Qwen3.8 Max 0902 yet, so there is no like-for-like coding score. On overall capability, Grok 4.7 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
GLM-5.2 and Qwen3.8 Max 0902 have the largest context windows (1,000,000 and 1,000,000 tokens), against 500,000 for Grok 4.7. Maximum output per response: GLM-5.2 up to 131,072, Grok 4.7 up to 500,000, Qwen3.8 Max 0902 up to 131,072 tokens.
Which can read images, PDFs, audio or video?
GLM-5.2 accepts text; Grok 4.7 accepts text, images and PDFs; Qwen3.8 Max 0902 accepts text, images, PDFs and video. Qwen3.8 Max 0902 handles the widest range of inputs.
Are any of these open source?
GLM-5.2 publishes its weights and can be self-hosted; Grok 4.7 and Qwen3.8 Max 0902 is proprietary.
Which is newer?
Grok 4.7 is the newest, released Sep 21, 2026. Qwen3.8 Max 0902 came out Sep 2, 2026; GLM-5.2 came out Jun 13, 2026. Knowledge cutoff: Grok 4.7 May 2026.
How do you decide the winner?
This rating uses capability (59%, independent benchmark results); price (29%, blended price per million tokens (3 input : 1 output), log scale); context window (12%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 22, 2026.This rating uses capability (83%, independent benchmark results); context window (17%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 22, 2026.