Skip to main content
BenchLM
Data

Claude Opus 5 vs Gemma 4 31B

Updated October 20, 2026. Rank says Claude Opus 5 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInAPI/MCP

Decision reading

Claude Opus 5 has the higher public point estimate, 80.11 versus 39.84. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 2 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Anthropic logo
Claude Opus 5

Anthropic

80.11/100

Supported · Public rank #8

90% interval 77.2–83.0

Model B
Google logo
Gemma 4 31B

Google

39.84/100

Estimated · Public rank #139

Conditional range 30.1–49.6

Shared results
2
Claude Opus 5 only
78
Gemma 4 31B only
9
Like-for-like categories
2 / 8
Supported: Claude Opus 5 · Estimated: Gemma 4 31B. Conditional ranges do not establish rank confidence.How the comparison works

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Coding work

    Code generation, repair, and software-engineering tasks

    Claude Opus 5

    Claude Opus 5 has the higher public coding point estimate, 70.8 to 31.7, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    Claude Opus 5

    Claude Opus 5 has the larger documented context window.

    Confidence: documented
Show secondary and unsupported calls

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    Gemma 4 31B is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.

    Confidence: limited

  • Chat turn cost

    1K fresh input + 500 output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

  • Repository review cost

    50K fresh input + 3K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

70.8Claude Opus 531.7Gemma 4 31B

Like-for-like · BenchAlign v5.8

Claude Opus 5 has the higher coding point estimate. Conditional score ranges do not establish rank confidence.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboardShortlist models for this job
Same basis rules as the category table below

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

2 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Coding

Like-for-like
Claude Opus 5
70.8
Supported · #9/152
Gemma 4 31B
31.7
Supported · #90/152
Basis
BenchAlign v5.8 lane · 18 vs 3 public rows
Reading
Claude Opus 5 leads

Knowledge

Like-for-like
Claude Opus 5
82.3
Supported · #7/187
Gemma 4 31B
40.9
Supported · #117/187
Basis
BenchAlign v5.8 lane · 19 vs 4 public rows
Reading
Claude Opus 5 leads

Agentic

Directional only
Claude Opus 5
78.4
Supported · #4/131
Gemma 4 31B
20.1
Estimated · #111/131
Basis
BenchAlign v5.8 lane · 21 vs 1 public rows
Reading
Directional only

Multimodal

Directional only
Claude Opus 5
92.2
#2/55
Gemma 4 31B
74.3
#26/55
Basis
Provisional lane · 1 vs 1 weighted rows
Reading
Directional only

Reasoning

Not comparable
Claude Opus 5
81.0
#8/27
Gemma 4 31B
65.8
Unranked · 3 rankable rows
Basis
Provisional lane · 2 vs 1 weighted rows
Reading
Not comparable

Multilingual

Not comparable
Claude Opus 5
Not ranked
Gemma 4 31B
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Claude Opus 5
Not ranked
Gemma 4 31B
92.9
#15/149
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Claude Opus 5
91.6
Unranked · 7 rankable rows
Gemma 4 31B
Not ranked
Basis
Provisional lane · 2 vs 0 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Claude Opus 5
$0.0175
Fits in one request
Gemma 4 31B
Self-hosted; infrastructure cost varies
Fits in one request

Gemma 4 31B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Claude Opus 5
$0.325
Fits in one request
Gemma 4 31B
Self-hosted; infrastructure cost varies
Fits in one request

Gemma 4 31B has no comparable published API token rate.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Claude Opus 5
$0.45
Fits in one request
Gemma 4 31B
Self-hosted; infrastructure cost varies
Fits in one request
Cached-input rate unavailable

Gemma 4 31B has no comparable published API token rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Claude Opus 5

$0.5 per 1M cached input tokens

Claude API pricing

Gemma 4 31B

No comparable hosted API rate

Reasoning profile

Claude Opus 5

Reasoning

Gemma 4 31B

Reasoning

Weight access

Claude Opus 5

Proprietary

Gemma 4 31B

Open Weight

License

Claude Opus 5

Proprietary

Gemma 4 31B

Open Weight

Release date

Claude Opus 5

2026-07-24

Gemma 4 31B

2026-04-02

If you already use one of these models

Deployment change
The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal
Claude Opus 5 has the higher public point estimate, 80.11 versus 39.84. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Claude Opus 5 has the larger documented window (1M).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Claude Opus 5 or Gemma 4 31B?

Claude Opus 5 has the higher public point estimate, 80.11 versus 39.84. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Claude Opus 5 or Gemma 4 31B?

Claude Opus 5 has the higher public coding point estimate, 70.8 to 31.7, with Supported evidence for both models. The conditional ranges do not establish rank confidence.

Which is better for agentic tasks, Claude Opus 5 or Gemma 4 31B?

Claude Opus 5 scores higher for agentic tasks on the public lane, 78.4 to 20.1. Gemma 4 31B is scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Claude Opus 5 or Gemma 4 31B?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, Claude Opus 5 or Gemma 4 31B?

Claude Opus 5 has the larger documented context window: 1M, compared with 256K.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

Claude Opus 5
API / mo$22,500
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Gemma 4 31B
API / mo$0
Self-host / mo$298
Break-even—
Model the full break-even

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence89 rows

Agentic

  • Terminal-Bench 3.0

    Claude Opus 542.7%
    Source
    Gemma 4 31B—

    Not directly comparable

  • BrowseComp

    Claude Opus 590.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • HLE w/ tools

    Claude Opus 564.7%
    Source
    Gemma 4 31B—

    Not directly comparable

  • DeepSearchQA

    Claude Opus 595.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • DRACO

    Claude Opus 588.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • BrowseComp (10-agent, prerelease)

    Claude Opus 593.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • OSWorld 2.0

    Claude Opus 570.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MCP Atlas

    Claude Opus 585.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MCP-Atlas claim coverage

    Claude Opus 589.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LAB all-pass (Anthropic harness)

    Claude Opus 523.58%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LAB criterion-pass (Anthropic harness)

    Claude Opus 593.74%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LAB all-pass (Harvey held-out)

    Claude Opus 511.7%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LAB criterion-pass (Harvey held-out)

    Claude Opus 594.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Toolathlon-Verified

    Claude Opus 580.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Toolathlon Verified Pass@3

    Claude Opus 587.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Toolathlon Verified Pass³

    Claude Opus 573.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Toolathlon Verified avg. turns

    Claude Opus 523.5 turns
    Source
    Gemma 4 31B—

    Not directly comparable

  • AutomationBench

    Claude Opus 526.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Terminal-Bench 2.1 (Vals)

    Claude Opus 584.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ApprenticeBench

    Claude Opus 536%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MazeBench

    Claude Opus 512%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Gert Labs

    Claude Opus 5—
    Gemma 4 31B35.26%
    Source

    Not directly comparable

Coding

  • Bug Hunt Bench

    Claude Opus 527.0 fixes
    Source
    Gemma 4 31B—

    Not directly comparable

  • SWE-bench Verified

    Claude Opus 596%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SWE-bench Pro

    Claude Opus 579.2%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SWE Multilingual

    Claude Opus 589.5%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SWE Multimodal

    Claude Opus 559.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • DeepSWE

    Claude Opus 568.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • FrontierCode 1.1 Main

    Claude Opus 553.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • FrontierCode 1.1 Extended

    Claude Opus 563.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • FrontierSWE v2

    Claude Opus 552.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ProgramBench (episode 1)

    Claude Opus 583.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ProgramBench

    Claude Opus 593.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • CursorBench 3.2

    Claude Opus 570.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • VulcanBench v3

    Claude Opus 587.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • VulcanBench CII v1

    Claude Opus 596.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LiveCodeBench (Vals)

    Claude Opus 589.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SWE-bench (Vals)

    Claude Opus 597.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • CursorBench 4.0

    Claude Opus 546.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • PostTrainBench v1.1

    Claude Opus 535.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LiveCodeBench v6

    Claude Opus 5—
    Gemma 4 31B80.0%
    Source

    Not directly comparable

  • SWE-Rebench

    Claude Opus 5—
    Gemma 4 31B41.6%
    Source

    Not directly comparable

  • React Native Evals

    Claude Opus 5—
    Gemma 4 31B75.2%
    Source

    Not directly comparable

Reasoning

  • ARC-AGI-1

    Claude Opus 597.50%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ARC-AGI-2

    Claude Opus 590.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ARC-AGI-3

    Claude Opus 530.2%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MRCRv2

    Claude Opus 5—
    Gemma 4 31B66.4%
    Source

    Not directly comparable

Multimodal

  • Chartography (no tools)

    Claude Opus 529.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Chartography (tools)

    Claude Opus 583.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • BenchCAD Vision2Code (no tools)

    Claude Opus 50.366
    Source
    Gemma 4 31B—

    Not directly comparable

  • BenchCAD Vision2Code (tools)

    Claude Opus 50.821
    Source
    Gemma 4 31B—

    Not directly comparable

  • GDP.pdf (no tools)

    Claude Opus 583.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • GDP.pdf (tools)

    Claude Opus 585.5%
    Source
    Gemma 4 31B—

    Not directly comparable

  • OfficeQA

    Claude Opus 578.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • OfficeQA Pro

    Claude Opus 566.9%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ParseBench

    Claude Opus 5—
    Gemma 4 31B62.40%
    Source

    Not directly comparable

  • MMMU-Pro

    Claude Opus 5—
    Gemma 4 31B76.9%
    Source

    Not directly comparable

Knowledge

  • HLE

    Claude Opus 564.7%
    Source
    Gemma 4 31B26.5%
    Source

    Claude Opus 5 leads this result

  • HLE w/o tools

    Claude Opus 556.3%
    Source
    Gemma 4 31B19.5%
    Source

    Claude Opus 5 leads this result

  • HLE-Verified

    Claude Opus 554.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • LABBench2

    Claude Opus 584.2%
    Source
    Gemma 4 31B—

    Not directly comparable

  • HealthBench (raw)

    Claude Opus 567.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • HealthBench (length-adjusted)

    Claude Opus 557.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • HealthBench Professional

    Claude Opus 559.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • HealthBench Professional (raw)

    Claude Opus 573.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • BioMysteryBench (human-solvable)

    Claude Opus 590.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • BioMysteryBench (human-difficult)

    Claude Opus 549.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SpatialBench Verified

    Claude Opus 572.5%
    Source
    Gemma 4 31B—

    Not directly comparable

  • SingleCellBench

    Claude Opus 560.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ProteinGym Hard

    Claude Opus 547.7%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Protein Design

    Claude Opus 542.5%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Organic chemistry V2

    Claude Opus 561.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Protocols (troubleshooting)

    Claude Opus 561.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • Protocols (understanding)

    Claude Opus 578.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • GPQA Diamond (Vals)

    Claude Opus 593.4%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MMLU-Pro (Vals)

    Claude Opus 591.6%
    Source
    Gemma 4 31B—

    Not directly comparable

  • GPQA

    Claude Opus 5—
    Gemma 4 31B84.3%
    Source

    Not directly comparable

  • MMLU-Pro

    Claude Opus 5—
    Gemma 4 31B85.2%
    Source

    Not directly comparable

Multilingual

  • GMMLU

    Claude Opus 592.5%
    Source
    Gemma 4 31B—

    Not directly comparable

  • MILU

    Claude Opus 592.1%
    Source
    Gemma 4 31B—

    Not directly comparable

  • INCLUDE

    Claude Opus 589.8%
    Source
    Gemma 4 31B—

    Not directly comparable

Instruction following

  • Gray Swan IPI (15 attempts)

    Claude Opus 54.6%
    Source
    Gemma 4 31B—

    Not directly comparable

Math

  • IMO 2026

    Claude Opus 542/42
    Source
    Gemma 4 31B—

    Not directly comparable

  • RiemannBench (no tools)

    Claude Opus 560.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • RiemannBench (tools)

    Claude Opus 579.0%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ArXivMath Jun. 2026 (no tools)

    Claude Opus 590.8%
    Source
    Gemma 4 31B—

    Not directly comparable

  • ArXivMath Jun. 2026 (tools)

    Claude Opus 591.3%
    Source
    Gemma 4 31B—

    Not directly comparable

  • FrontierMath v2 (Tiers 1-3)

    Claude Opus 585.614%
    Source
    Gemma 4 31B—

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Claude Opus 573.171%
    Source
    Gemma 4 31B—

    Not directly comparable

89 public results · 2 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch Claude Opus 5 vs Gemma 4 31B

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 5,600+ readers.

Last updated October 20, 2026