GLM-5.1 vs Grok 4.20 (Non-Reasoning) vs DeepSeek V4 Pro
No independent capability benchmark covers every model here yet. The score uses price and context window.No independent capability benchmark covers every model here yet. The score uses context window.
Z.ai (Zhipu)
GLM-5.1
34/100- ECI150.7
- Price$1.40 / $4.40
- Context200K
xAI
Grok 4.20 (Non-Reasoning)
46/100- ECI—
- Price$1.25 / $2.50
- Context1M
DeepSeek
DeepSeek V4 Pro
42/100- ECI150.2
- Price$1.60 / $3.33
- Context1M
Z.ai (Zhipu)
GLM-5.1
39/100- ECI150.7
- Price$1.40 / $4.40
- Context200K
xAI
Grok 4.20 (Non-Reasoning)
64/100- ECI—
- Price$1.25 / $2.50
- Context1M
DeepSeek
DeepSeek V4 Pro
64/100- ECI150.2
- Price$1.60 / $3.33
- Context1M
Capability evidence is incomplete
There is not enough shared capability evidence to name an overall winner. The scores (Grok 4.20 (Non-Reasoning) 46/100, DeepSeek V4 Pro 42/100, GLM-5.1 34/100) compare the available specifications and price. The score weighs price 71%, context window 29%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.There is not enough shared capability evidence to name an overall winner. The scores (Grok 4.20 (Non-Reasoning) 64/100, DeepSeek V4 Pro 64/100, GLM-5.1 39/100) compare the available specifications only. The score weighs context window 100%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
- CapabilityNot enough dataNo independent benchmark covers every model here yet
- Lowest priceGrok 4.20 (Non-Reasoning)Grok 4.20 (Non-Reasoning) $1.56 · DeepSeek V4 Pro $2.03 · GLM-5.1 $2.15 per 1M tokens (3:1 blend)
- Longest contextGrok 4.20 (Non-Reasoning) and DeepSeek V4 ProGrok 4.20 (Non-Reasoning) 1,000,000 · DeepSeek V4 Pro 1,000,000 · GLM-5.1 200,000 tokens
- Widest inputsGrok 4.20 (Non-Reasoning)GLM-5.1: Text · Grok 4.20 (Non-Reasoning): Text, Images, PDFs · DeepSeek V4 Pro: Text
- Self-hostingGLM-5.1 and DeepSeek V4 ProPublishes downloadable weights
| Measure | Weight | GLM-5.1 | Grok 4.20 (Non-Reasoning) | DeepSeek V4 Pro |
|---|---|---|---|---|
| Price | 71% | 33 | 39 | 34 |
| Context window | 29% | 39 | 64 | 64 |
| Specs & price only | 100% | 34/100 | 46/100 | 42/100 |
Left out because at least one model lacks the data: capability and inputs & features. The remaining weights were rescaled.
| Measure | Weight | GLM-5.1 | Grok 4.20 (Non-Reasoning) | DeepSeek V4 Pro |
|---|---|---|---|---|
| Context window | 100% | 39 | 64 | 64 |
| Specs only | 100% | 39/100 | 64/100 | 64/100 |
Left out because at least one model lacks the data: capability and 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) | 150.7 (best) | Not published | 150.2 |
| ECI rank | #67 of 159 (best) | Not published | #68 of 159 |
| ECI confidence interval | 148.7–152.8 | Not published | 148.4–152.0 |
| ECI evaluation date | Apr 7, 2026 | Not published | Apr 24, 2026 |
| Documented capabilities | Strong GLM coding model for agentic engineering, terminals, and repository generation | Grok model for agentic tool use, reasoning, coding, and live assistance | Open MoE flagship with million-token context for coding and long agent runs |
| Model parameters | Not published | Not published | Not published |
| Output modalities | Text | Text | Text |
| GPQA DiamondGraduate-level science questions | 89.9% | Not published | 90.9%high (best) |
| FrontierMath Tiers 1–3Research-level mathematics | 36.8% | Not published | 45.3%max (best) |
| OTIS Mock AIME 2024–2025Competition mathematics | 93.3% | Not published | 96.7%max (best) |
| SWE-bench VerifiedFixing real GitHub issues | 74.2% | Not published | 77.6%max (best) |
| SimpleQA VerifiedShort factual questions | 34.0% | Not published | 47.0%max (best) |
| Publisher-reported benchmarks | |||
| Artificial Analysis Coding Agent IndexPublisher-reported result | 52.7average pass@1Reported result | Not published | Not published |
| SWE-Atlas Codebase QnAPublisher-reported result | 73.2pass@1Reported result | Not published | Not published |
| SWE-Bench Pro · hard-aaPublisher-reported result | 19.8pass@1Reported result | Not published | Not published |
| Terminal-Bench 2.1Publisher-reported result | 65.1pass@1Reported result | Not published | Not published |
| SWE-Bench VerifiedPublisher-reported result | Not published | Not published | 80.6resolvedReported result |
| Artificial Analysis Coding Agent Indexhigh | Not published | Not published | 50.1average pass@1Reported result |
| SWE-Atlas Codebase QnAhigh | Not published | Not published | 67.8pass@1Reported result |
| SWE-Bench Pro · hard-aahigh | Not published | Not published | 18pass@1Reported result |
| Terminal-Bench 2.1high | Not published | Not published | 64.7pass@1Reported result |
| MMLU-Propreview checkpoint; max effort | Not published | Not published | 87.5EMReported result |
| SimpleQA-Verifiedpreview checkpoint; max effort | Not published | Not published | 57.9pass@1Reported result |
| Chinese SimpleQApreview checkpoint; max effort | Not published | Not published | 84.4pass@1Reported result |
| GPQA Diamondpreview checkpoint; max effort | Not published | Not published | 90.1pass@1Reported result |
| Humanity's Last Exampreview checkpoint; max effort; without tools | Not published | Not published | 37.7pass@1Reported result |
| LiveCodeBenchpreview checkpoint; max effort | Not published | Not published | 93.5pass@1Reported result |
| Codeforcespreview checkpoint; max effort | Not published | Not published | 3206ratingReported result |
| HMMT · February 2026preview checkpoint; max effort | Not published | Not published | 95.2pass@1Reported result |
| IMOAnswerBenchpreview checkpoint; max effort | Not published | Not published | 89.8pass@1Reported result |
| MathArena Apexpreview checkpoint; max effort | Not published | Not published | 38.3pass@1Reported result |
| MathArena Apex Shortlistpreview checkpoint; max effort | Not published | Not published | 90.2pass@1Reported result |
| MRCR · 1M contextpreview checkpoint; max effort | Not published | Not published | 83.5MMRReported result |
| CorpusQA · 1M contextpreview checkpoint; max effort | Not published | Not published | 62accuracyReported result |
| Terminal-Bench 2.0preview checkpoint; max effort | Not published | Not published | 67.9accuracyReported result |
| SWE-Bench Propreview checkpoint; max effort | Not published | Not published | 55.4resolvedReported result |
| SWE-Bench Multilingualpreview checkpoint; max effort | Not published | Not published | 76.2resolvedReported result |
| BrowseComppreview checkpoint; max effort | Not published | Not published | 83.4pass@1Reported result |
| Humanity's Last Exampreview checkpoint; max effort; with tools | Not published | Not published | 48.2pass@1Reported result |
| MCP Atlas · publicpreview checkpoint; max effort | Not published | Not published | 73.6pass@1Reported result |
| GDPval-AApreview checkpoint; max effort | Not published | Not published | 1554EloReported result |
| Toolathlonpreview checkpoint; max effort | Not published | Not published | 51.8pass@1Reported result |
| Price per million tokens | |||
| Input | $1.40 | $1.25 (best) | $1.60 |
| Output | $4.40 | $2.50 (best) | $3.33 |
| Cached input | $0.26 | $0.20 (best) | Not published |
| Blended (3:1) | $2.15 | $1.56 (best) | $2.03 |
| Long-context rate | Same rate | Over 200K: $2.50 / $5.00 | Same rate |
| Price source | Official Z.AI API | Official xAI API | Median of 65 providers |
| Limits | |||
| Context window | 200,000 tokens | 1,000,000 tokens (best) | 1,000,000 tokens (best) |
| Max input | Not published | Not published | Not published |
| Max output | 131,072 tokens | 30,000 tokens | 384,000 tokens (best) |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | Yes | No |
| PDFs | No | Yes | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yes | No | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | Not published |
| File attachments | No | Yes | No |
| Temperature control | Yes | Yes | Yes |
| Reasoning control | toggle | Not published | always |
| Availability | |||
| Weights | Open | Proprietary | Open |
| API model ID | glm-5.1 | grok-4.20-0309-non-reasoning | Not published |
| API providers | 45 | 13 | 67 (best) |
| Released | Apr 7, 2026 | Mar 9, 2026 | Apr 24, 2026 |
| Knowledge cutoff | Not published | Not published | May 2025 |
| Publisher documentation | Publisher documentation · Model weights | Publisher documentation | Model weights |
| Source metadata updated | Apr 7, 2026 | Mar 9, 2026 | Apr 24, 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-5.1$22.80
Grok 4.20 (Non-Reasoning)$17.50
DeepSeek V4 Pro$22.66
Which should you choose?
Which is better: GLM-5.1, Grok 4.20 (Non-Reasoning) or DeepSeek V4 Pro?
There is not enough shared capability evidence to name an overall winner. The scores (Grok 4.20 (Non-Reasoning) 46/100, DeepSeek V4 Pro 42/100, GLM-5.1 34/100) compare the available specifications and price. The score weighs price 71%, context window 29%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.There is not enough shared capability evidence to name an overall winner. The scores (Grok 4.20 (Non-Reasoning) 64/100, DeepSeek V4 Pro 64/100, GLM-5.1 39/100) compare the available specifications only. The score weighs context window 100%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.
Which is cheaper, GLM-5.1, Grok 4.20 (Non-Reasoning) or DeepSeek V4 Pro?
Grok 4.20 (Non-Reasoning) is cheaper at $1.25 input / $2.50 output per million tokens (official xAI API price). DeepSeek V4 Pro costs $1.60 input / $3.33 output per million tokens (median across 65 API providers); GLM-5.1 costs $1.40 input / $4.40 output per million tokens (official Z.AI API price). At a typical mix of three input tokens to one output token, that is $1.56 per million tokens for Grok 4.20 (Non-Reasoning) versus $2.03 for DeepSeek V4 Pro (1.3× as much) and $2.15 for GLM-5.1 (1.4× as much).
Which scores higher on benchmarks?
There is no independent benchmark that covers all three models yet. GLM-5.1 has an ECI of 150.7, Grok 4.20 (Non-Reasoning) has not been scored yet and DeepSeek V4 Pro has an ECI of 150.2.
Which is better for coding?
There are no published SWE-bench Verified results for Grok 4.20 (Non-Reasoning) yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.
Which has the bigger context window?
Grok 4.20 (Non-Reasoning) and DeepSeek V4 Pro have the largest context windows (1,000,000 and 1,000,000 tokens), against 200,000 for GLM-5.1. Maximum output per response: GLM-5.1 up to 131,072, Grok 4.20 (Non-Reasoning) up to 30,000, DeepSeek V4 Pro up to 384,000 tokens.
Which can read images, PDFs, audio or video?
GLM-5.1 accepts text; Grok 4.20 (Non-Reasoning) accepts text, images and PDFs; DeepSeek V4 Pro accepts text. Grok 4.20 (Non-Reasoning) handles the widest range of inputs.
Are any of these open source?
GLM-5.1 and DeepSeek V4 Pro publishes its weights and can be self-hosted; Grok 4.20 (Non-Reasoning) is proprietary.
Which is newer?
DeepSeek V4 Pro is the newest, released Apr 24, 2026. GLM-5.1 came out Apr 7, 2026; Grok 4.20 (Non-Reasoning) came out Mar 9, 2026. Knowledge cutoff: DeepSeek V4 Pro May 2025.
How do you decide the winner?
This rating uses price (71%, blended price per million tokens (3 input : 1 output), log scale); context window (29%, 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 context window (100%, 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.