Skip to main content
BenchLM
Data

Gemini 4 Carbon vs Step 5 Preview

Updated October 28, 2026. Rank says Gemini 4 Carbon 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

Gemini 4 Carbon has the higher public point estimate, 88.42 versus 56.2. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 1 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Google logo
Gemini 4 Carbon

Google

88.42/100

Estimated · Public rank #1

Conditional range 74.1–100.0

Model B
StepFun logo
Step 5 Preview

StepFun

56.2/100

Estimated · Public rank #61

Conditional range 46.5–65.9

Shared results
1
Gemini 4 Carbon only
8
Step 5 Preview only
23
Like-for-like categories
1 / 8
Estimated: Gemini 4 Carbon and Step 5 Preview. 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.

  • Chat turn cost

    1K fresh input + 500 output tokens

    Step 5 Preview

    Step 5 Preview has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

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

    Step 5 Preview

    Step 5 Preview has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates
Show secondary and unsupported calls

  • Repository review cost

    50K fresh input + 3K output tokens

    Step 5 Preview

    Step 5 Preview has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    Step 5 Preview is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like.

    Confidence: limited

  • Agentic work

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

    Not enough matched evidence

    Gemini 4 Carbon is not ranked on the public lane for agentic, so no winner is named for agentic.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    Not enough matched evidence

    A complete context comparison is not sourced.

    Confidence: limited

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.

82.2Gemini 4 Carbon48.6Step 5 Preview

Directional only · BenchAlign v5.8

Gemini 4 Carbon 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.

1 category rests 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.

Knowledge

Like-for-like
Gemini 4 Carbon
88.2
Supported · #2/178
Step 5 Preview
59.8
Supported · #45/178
Basis
BenchAlign v5.8 lane · 0 vs 2 public rows
Reading
Gemini 4 Carbon leads

Coding

Directional only
Gemini 4 Carbon
82.2
Supported · #3/156
Step 5 Preview
48.6
Estimated · #49/156
Basis
BenchAlign v5.8 lane · 1 vs 6 public rows
Reading
Directional only

Agentic

Not comparable
Gemini 4 Carbon
Not ranked
Step 5 Preview
54.7
Estimated · #41/134
Basis
BenchAlign v5.8 lane · 4 vs 13 public rows
Reading
Not comparable

Reasoning

Not comparable
Gemini 4 Carbon
79.5
Unranked · 2 rankable rows
Step 5 Preview
82.5
Unranked · 2 rankable rows
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Multimodal

Not comparable
Gemini 4 Carbon
Not ranked
Step 5 Preview
73.4
#32/58
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Not comparable

Multilingual

Not comparable
Gemini 4 Carbon
Not ranked
Step 5 Preview
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Gemini 4 Carbon
Not ranked
Step 5 Preview
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Gemini 4 Carbon
Not ranked
Step 5 Preview
Not ranked
Basis
Provisional lane · 0 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

Gemini 4 Carbon
$0.007
Fit state unavailable
Step 5 Preview
$0.00235
Fits in one request

Step 5 Preview has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

Gemini 4 Carbon
$0.13
Fit state unavailable
Step 5 Preview
$0.0581
Fits in one request

Step 5 Preview has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

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

Gemini 4 Carbon
$0.16
Fit state unavailable
Step 5 Preview
$0.057
Fits in one request

Step 5 Preview has the lower modeled cost

Costs use the listed standard API rates.

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.

Reasoning profile

Gemini 4 Carbon

Reasoning

Step 5 Preview

Reasoning

Weight access

Gemini 4 Carbon

Proprietary

Step 5 Preview

Open Weight

License

Gemini 4 Carbon

Proprietary

Step 5 Preview

Open Weight

Release date

Gemini 4 Carbon

2026-10-28

Step 5 Preview

2026-09-20

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
Gemini 4 Carbon has the higher public point estimate, 88.42 versus 56.2. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
Repository review: $0.13 vs $0.0581. Cache-heavy agent loop: $0.16 vs $0.057.
Context tradeoff
A complete documented context comparison is not available.
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Gemini 4 Carbon or Step 5 Preview?

Gemini 4 Carbon has the higher public point estimate, 88.42 versus 56.2. 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, Gemini 4 Carbon or Step 5 Preview?

Gemini 4 Carbon scores higher for coding on the public lane, 82.2 to 48.6. Step 5 Preview is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like. BenchLM does not name a winner for coding from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which is better for agentic tasks, Gemini 4 Carbon or Step 5 Preview?

Gemini 4 Carbon is not ranked on the public lane for agentic tasks, so no winner is named for agentic tasks.

Which costs less, Gemini 4 Carbon or Step 5 Preview?

For the stated presets, chat costs $0.007 on Gemini 4 Carbon and $0.00235 on Step 5 Preview; repository review costs $0.13 and $0.0581; the cache-heavy agent loop costs $0.16 and $0.057. Costs use the listed standard API rates.

Which has the larger context window, Gemini 4 Carbon or Step 5 Preview?

A complete documented context-window comparison is not available.

Self-host vs API cost

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

Gemini 4 Carbon
API / mo$9,000
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Step 5 Preview
API / mo$2,775
Self-host / mo$27,962
Break-even617M/day
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 evidence32 rows

Agentic

  • AutomationBench

    Gemini 4 Carbon53.9%
    Source
    Step 5 Preview44.0%
    Source

    Gemini 4 Carbon leads this result

  • OSWorld 2.1 Offline

    Gemini 4 Carbon81.5%
    Source
    Step 5 Preview—

    Not directly comparable

  • CWE-bench v1

    Gemini 4 Carbon71.0%
    Source
    Step 5 Preview—

    Not directly comparable

  • ExploitGym

    Gemini 4 Carbon53.5%
    Source
    Step 5 Preview—

    Not directly comparable

  • Terminal-Bench 2.1

    Gemini 4 Carbon—
    Step 5 Preview85.0%
    Source

    Not directly comparable

  • Terminal-Bench 4.0

    Gemini 4 Carbon—
    Step 5 Preview33.30%
    Source

    Not directly comparable

  • CyberGym

    Gemini 4 Carbon—
    Step 5 Preview84.7%
    Source

    Not directly comparable

  • Toolathlon-Verified

    Gemini 4 Carbon—
    Step 5 Preview74.1%
    Source

    Not directly comparable

  • MCP Atlas

    Gemini 4 Carbon—
    Step 5 Preview85.6%
    Source

    Not directly comparable

  • JobBench

    Gemini 4 Carbon—
    Step 5 Preview59.0%
    Source

    Not directly comparable

  • APEX-Agents

    Gemini 4 Carbon—
    Step 5 Preview37.8%
    Source

    Not directly comparable

  • DRACO

    Gemini 4 Carbon—
    Step 5 Preview83.3%
    Source

    Not directly comparable

  • BrowseComp

    Gemini 4 Carbon—
    Step 5 Preview88.7%
    Source

    Not directly comparable

  • Agents' Last Exam

    Gemini 4 Carbon—
    Step 5 Preview29.5%
    Source

    Not directly comparable

  • Hermes Index

    Gemini 4 Carbon—
    Step 5 Preview32.8%
    Source

    Not directly comparable

  • Hermes Bench

    Gemini 4 Carbon—
    Step 5 Preview69.2%
    Source

    Not directly comparable

Coding

  • FrontierSWE v2

    Gemini 4 Carbon61.2%
    Source
    Step 5 Preview—

    Not directly comparable

  • DeepSWE

    Gemini 4 Carbon—
    Step 5 Preview67.7%
    Source

    Not directly comparable

  • Terminal-Bench 2.1

    Gemini 4 Carbon—
    Step 5 Preview85.0%
    Source

    Not directly comparable

  • SciCode

    Gemini 4 Carbon—
    Step 5 Preview58.9%
    Source

    Not directly comparable

  • ProgramBench

    Gemini 4 Carbon—
    Step 5 Preview80.5%
    Source

    Not directly comparable

  • sweMarathon

    Gemini 4 Carbon—
    Step 5 Preview72.7%
    Source

    Not directly comparable

  • MLS-Bench Lite

    Gemini 4 Carbon—
    Step 5 Preview40.5%
    Source

    Not directly comparable

Reasoning

  • GraphWalks BFS 256K–1M

    Gemini 4 Carbon96.2%
    Source
    Step 5 Preview—

    Not directly comparable

  • CritPt

    Gemini 4 Carbon—
    Step 5 Preview20.9%
    Source

    Not directly comparable

Multimodal

  • Chartography (no tools)

    Gemini 4 Carbon72.5%
    Source
    Step 5 Preview—

    Not directly comparable

  • Blueprint-Bench 2

    Gemini 4 Carbon58.6%
    Source
    Step 5 Preview—

    Not directly comparable

  • OfficeQA Pro

    Gemini 4 Carbon—
    Step 5 Preview60.3%
    Source

    Not directly comparable

  • MMMU-Pro

    Gemini 4 Carbon—
    Step 5 Preview76%
    Source

    Not directly comparable

Knowledge

  • GPQA-D

    Gemini 4 Carbon—
    Step 5 Preview93.5%
    Source

    Not directly comparable

  • HLE

    Gemini 4 Carbon—
    Step 5 Preview46.5%
    Source

    Not directly comparable

Instruction following

  • Gray Swan IPI (15 attempts)

    Gemini 4 Carbon0.0%
    Source
    Step 5 Preview—

    Not directly comparable

32 public results · 1 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch Gemini 4 Carbon vs Step 5 Preview

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

Read a sample issue

Join 6,100+ subscribers.

Last updated October 28, 2026