ZMIME
Comparison · 3 models · Updated Nov 22, 2026

DeepSeek V4 Pro vs GLM-5.1 vs Grok 4.20 (Non-Reasoning)

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.

  1. DeepSeek

    DeepSeek V4 Pro

    Released Apr 24, 2026

    42/100
    • ECI150.2
    • Price$1.60 / $3.33
    • Context1M
  2. Z.ai (Zhipu)

    GLM-5.1

    Released Apr 7, 2026

    34/100
    • ECI150.7
    • Price$1.40 / $4.40
    • Context200K
  3. xAI

    Grok 4.20 (Non-Reasoning)

    Released Mar 9, 2026

    46/100
    • ECI—
    • Price$1.25 / $2.50
    • Context1M
  4. DeepSeek

    DeepSeek V4 Pro

    Released Apr 24, 2026

    64/100
    • ECI150.2
    • Price$1.60 / $3.33
    • Context1M
  5. Z.ai (Zhipu)

    GLM-5.1

    Released Apr 7, 2026

    39/100
    • ECI150.7
    • Price$1.40 / $4.40
    • Context200K
  6. xAI

    Grok 4.20 (Non-Reasoning)

    Released Mar 9, 2026

    64/100
    • ECI—
    • Price$1.25 / $2.50
    • Context1M
01 — Verdict

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 (DeepSeek V4 Pro 64/100, Grok 4.20 (Non-Reasoning) 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 contextDeepSeek V4 Pro and Grok 4.20 (Non-Reasoning)DeepSeek V4 Pro 1,000,000 · Grok 4.20 (Non-Reasoning) 1,000,000 · GLM-5.1 200,000 tokens
  • Widest inputsGrok 4.20 (Non-Reasoning)DeepSeek V4 Pro: Text · GLM-5.1: Text · Grok 4.20 (Non-Reasoning): Text, Images, PDFs
  • Self-hostingDeepSeek V4 Pro and GLM-5.1Publishes downloadable weights
How the score is built
MeasureWeightDeepSeek V4 ProGLM-5.1Grok 4.20 (Non-Reasoning)
Price71%343339
Context window29%643964
Specs & price only100%42/10034/10046/100

Left out because at least one model lacks the data: capability and inputs & features. The remaining weights were rescaled.

How the score is built
MeasureWeightDeepSeek V4 ProGLM-5.1Grok 4.20 (Non-Reasoning)
Context window100%643964
Specs only100%64/10039/10064/100

Left out because at least one model lacks the data: capability and inputs & features. The remaining weights were rescaled.

02 — Side by side

Every spec in one table

Highlighted cells lead their row. Unavailable specifications are marked “Not published”. Publisher results include their evaluation settings and source.

DeepSeek V4 Pro vs GLM-5.1 vs Grok 4.20 (Non-Reasoning) specifications side by side
SpecificationDeepSeek V4 ProDeepSeekGLM-5.1Z.ai (Zhipu)Grok 4.20 (Non-Reasoning)xAI
Capability
Capabilities Index (ECI)150.2150.7 (best)Not published
ECI rank#68 of 159#67 of 159 (best)Not published
ECI confidence interval148.4–152.0148.7–152.8Not published
ECI evaluation dateApr 24, 2026Apr 7, 2026Not published
Documented capabilitiesOpen MoE flagship with million-token context for coding and long agent runsStrong GLM coding model for agentic engineering, terminals, and repository generationGrok model for agentic tool use, reasoning, coding, and live assistance
Model parametersNot publishedNot publishedNot published
Output modalitiesTextTextText
GPQA DiamondGraduate-level science questions90.9%high (best)89.9%Not published
FrontierMath Tiers 1–3Research-level mathematics45.3%max (best)36.8%Not published
OTIS Mock AIME 2024–2025Competition mathematics96.7%max (best)93.3%Not published
SWE-bench VerifiedFixing real GitHub issues77.6%max (best)74.2%Not published
SimpleQA VerifiedShort factual questions47.0%max (best)34.0%Not published
Publisher-reported benchmarks
SWE-Bench VerifiedPublisher-reported result80.6resolvedReported resultNot publishedNot published
Artificial Analysis Coding Agent Indexhigh50.1average pass@1Reported resultNot publishedNot published
SWE-Atlas Codebase QnAhigh67.8pass@1Reported resultNot publishedNot published
SWE-Bench Pro · hard-aahigh18pass@1Reported resultNot publishedNot published
Terminal-Bench 2.1high64.7pass@1Reported resultNot publishedNot published
MMLU-Propreview checkpoint; max effort87.5EMReported resultNot publishedNot published
SimpleQA-Verifiedpreview checkpoint; max effort57.9pass@1Reported resultNot publishedNot published
Chinese SimpleQApreview checkpoint; max effort84.4pass@1Reported resultNot publishedNot published
GPQA Diamondpreview checkpoint; max effort90.1pass@1Reported resultNot publishedNot published
Humanity's Last Exampreview checkpoint; max effort; without tools37.7pass@1Reported resultNot publishedNot published
LiveCodeBenchpreview checkpoint; max effort93.5pass@1Reported resultNot publishedNot published
Codeforcespreview checkpoint; max effort3206ratingReported resultNot publishedNot published
HMMT · February 2026preview checkpoint; max effort95.2pass@1Reported resultNot publishedNot published
IMOAnswerBenchpreview checkpoint; max effort89.8pass@1Reported resultNot publishedNot published
MathArena Apexpreview checkpoint; max effort38.3pass@1Reported resultNot publishedNot published
MathArena Apex Shortlistpreview checkpoint; max effort90.2pass@1Reported resultNot publishedNot published
MRCR · 1M contextpreview checkpoint; max effort83.5MMRReported resultNot publishedNot published
CorpusQA · 1M contextpreview checkpoint; max effort62accuracyReported resultNot publishedNot published
Terminal-Bench 2.0preview checkpoint; max effort67.9accuracyReported resultNot publishedNot published
SWE-Bench Propreview checkpoint; max effort55.4resolvedReported resultNot publishedNot published
SWE-Bench Multilingualpreview checkpoint; max effort76.2resolvedReported resultNot publishedNot published
BrowseComppreview checkpoint; max effort83.4pass@1Reported resultNot publishedNot published
Humanity's Last Exampreview checkpoint; max effort; with tools48.2pass@1Reported resultNot publishedNot published
MCP Atlas · publicpreview checkpoint; max effort73.6pass@1Reported resultNot publishedNot published
GDPval-AApreview checkpoint; max effort1554EloReported resultNot publishedNot published
Toolathlonpreview checkpoint; max effort51.8pass@1Reported resultNot publishedNot published
Artificial Analysis Coding Agent IndexPublisher-reported resultNot published52.7average pass@1Reported resultNot published
SWE-Atlas Codebase QnAPublisher-reported resultNot published73.2pass@1Reported resultNot published
SWE-Bench Pro · hard-aaPublisher-reported resultNot published19.8pass@1Reported resultNot published
Terminal-Bench 2.1Publisher-reported resultNot published65.1pass@1Reported resultNot published
Price per million tokens
Input$1.60$1.40$1.25 (best)
Output$3.33$4.40$2.50 (best)
Cached inputNot published$0.26$0.20 (best)
Blended (3:1)$2.03$2.15$1.56 (best)
Long-context rateSame rateSame rateOver 200K: $2.50 / $5.00
Price sourceMedian of 65 providersOfficial Z.AI APIOfficial xAI API
Limits
Context window1,000,000 tokens (best)200,000 tokens1,000,000 tokens (best)
Max inputNot publishedNot publishedNot published
Max output384,000 tokens (best)131,072 tokens30,000 tokens
Inputs and features
TextYesYesYes
ImagesNoNoYes
PDFsNoNoYes
AudioNoNoNo
VideoNoNoNo
ReasoningYesYesNo
Tool callingYesYesYes
Structured outputNot publishedYesYes
File attachmentsNoNoYes
Temperature controlYesYesYes
Reasoning controlalwaystoggleNot published
Availability
WeightsOpenOpenProprietary
API model IDNot publishedglm-5.1grok-4.20-0309-non-reasoning
API providers67 (best)4513
ReleasedApr 24, 2026Apr 7, 2026Mar 9, 2026
Knowledge cutoffMay 2025Not publishedNot published
Publisher documentationModel weightsPublisher documentation · Model weightsPublisher documentation
Source metadata updatedApr 24, 2026Apr 7, 2026Mar 9, 2026
Data sourcesmodels.dev · Epoch AImodels.dev · Epoch AImodels.dev · Epoch AI
Catalog checkedNov 22, 2026Nov 22, 2026Nov 22, 2026
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • DeepSeek V4 Pro$22.66
  • GLM-5.1$22.80
  • Grok 4.20 (Non-Reasoning)$17.50
04 — Questions

Which should you choose?

Which is better: DeepSeek V4 Pro, GLM-5.1 or Grok 4.20 (Non-Reasoning)?

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 (DeepSeek V4 Pro 64/100, Grok 4.20 (Non-Reasoning) 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, DeepSeek V4 Pro, GLM-5.1 or Grok 4.20 (Non-Reasoning)?

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. DeepSeek V4 Pro has an ECI of 150.2, GLM-5.1 has an ECI of 150.7 and Grok 4.20 (Non-Reasoning) has not been scored yet.

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?

DeepSeek V4 Pro and Grok 4.20 (Non-Reasoning) have the largest context windows (1,000,000 and 1,000,000 tokens), against 200,000 for GLM-5.1. Maximum output per response: DeepSeek V4 Pro up to 384,000, GLM-5.1 up to 131,072, Grok 4.20 (Non-Reasoning) up to 30,000 tokens.

Which can read images, PDFs, audio or video?

DeepSeek V4 Pro accepts text; GLM-5.1 accepts text; Grok 4.20 (Non-Reasoning) accepts text, images and PDFs. Grok 4.20 (Non-Reasoning) handles the widest range of inputs.

Are any of these open source?

DeepSeek V4 Pro and GLM-5.1 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.