Long documents
Prompts that approach the documented context limit
Kimi K3
Kimi K3 has the larger documented context window.
Confidence: documented
Kimi K3 vs Step 5 Preview
Updated November 5, 2026. Rank says Kimi K3 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.
Share or export
Decision reading
Kimi K3 has the higher public point estimate, 69.46 versus 56.81. Their conditional score ranges overlap. These ranges do not establish rank confidence. 17 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
- Shared results
- 17
- Kimi K3 only
- 41
- Step 5 Preview only
- 7
- Like-for-like categories
- 1 / 8
Supported: Kimi K3 · Estimated: 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-ratesCache-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-ratesCoding 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: limitedAgentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
Step 5 Preview is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.
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.
58.6Kimi K348.8Step 5 Preview
Directional only · BenchAlign v5.8
Kimi K3 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.
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.
41 Kimi K3 only
17 Shared
7 Step 5 Preview only
3 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.
AutomationBenchAgentic
Normalized gap 13.2Kimi K3:30.8%Step 5 Preview:44.0%HLEKnowledge
Normalized gap 9.5Kimi K3:56%Step 5 Preview:46.5%MMMU-ProMultimodal
Normalized gap 5.6Kimi K3:81.6%Step 5 Preview:76%OfficeQA ProMultimodal
Normalized gap 3.0Kimi K3:63.3%Step 5 Preview:60.3%BrowseCompAgentic
Normalized gap 2.5Kimi K3:91.2%Step 5 Preview:88.7%
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- Kimi K3
- 67.4
- Supported · #23/177
- Step 5 Preview
- 61.0
- Supported · #44/177
- Basis
- BenchAlign v5.8 lane · 6 vs 2 public rows
- Reading
- Kimi K3 leads · intervals overlap
Agentic
Directional only- Kimi K3
- 68.0
- Supported · #14/135
- Step 5 Preview
- 55.3
- Estimated · #42/135
- Basis
- BenchAlign v5.8 lane · 18 vs 13 public rows
- Reading
- Directional only
Coding
Directional only- Kimi K3
- 58.6
- Supported · #20/156
- Step 5 Preview
- 48.8
- Estimated · #47/156
- Basis
- BenchAlign v5.8 lane · 17 vs 6 public rows
- Reading
- Directional only
Multimodal
Directional only- Kimi K3
- 90.0
- #6/58
- Step 5 Preview
- 73.3
- #32/58
- Basis
- Provisional lane · 3 vs 2 weighted rows
- Reading
- Directional only
Reasoning
Not comparable- Kimi K3
- 69.7
- #15/26
- Step 5 Preview
- 82.5
- Unranked · 2 rankable rows
- Basis
- Provisional lane · 1 vs 0 weighted rows
- Reading
- Not comparable
Multilingual
Not comparable- Kimi K3
- Not ranked
- Step 5 Preview
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Instruction following
Not comparable- Kimi K3
- Not ranked
- Step 5 Preview
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Math
Not comparable- Kimi K3
- Not ranked
- Step 5 Preview
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
| Category | Kimi K3 | Step 5 Preview | Basis | Reading |
|---|---|---|---|---|
| Knowledge | 67.4Supported · #23/177 | 61.0Supported · #44/177 | Like-for-likeBenchAlign v5.8 lane · 6 vs 2 public rows | Kimi K3 leads · intervals overlap |
| Agentic | 68.0Supported · #14/135 | 55.3Estimated · #42/135 | Directional onlyBenchAlign v5.8 lane · 18 vs 13 public rows | Directional only |
| Coding | 58.6Supported · #20/156 | 48.8Estimated · #47/156 | Directional onlyBenchAlign v5.8 lane · 17 vs 6 public rows | Directional only |
| Multimodal | 90.0#6/58 | 73.3#32/58 | Directional onlyProvisional lane · 3 vs 2 weighted rows | Directional only |
| Reasoning | 69.7#15/26 | 82.5Unranked · 2 rankable rows | Not comparableProvisional lane · 1 vs 0 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Instruction following | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | 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
- Kimi K3
- $0.0105
- Fits in one request
- 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
- Kimi K3
- $0.195
- Fits in one request
- 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
- Kimi K3
- $0.27
- Fits in one request
- 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.
SpecificationKimi K3Step 5 Preview
Context window
Maximum documented context; output-token limits may be lower.
Kimi K3
1.05M
Step 5 Preview
API model ID
Kimi K3
Not sourced
Step 5 Preview
step-5-preview
StepFun Step 5 Preview model documentationCached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Kimi K3
$0.3 per 1M cached input tokens
Step 5 Preview
$0.05 per 1M cached input tokens
StepFun pricing and rate limitsDocumented inputs
Kimi K3
Not sourced
Step 5 Preview
text, image, video
StepFun Step 5 Preview model documentationDocumented outputs
Kimi K3
Not sourced
Step 5 Preview
Provider availability
Kimi K3
Not sourced
Step 5 Preview
Preview · StepFun Open Platform API, Hugging Face, self-hosted inference
StepFun Step 5 Preview model documentationReasoning profile
Kimi K3
Reasoning
Step 5 Preview
Reasoning
Weight access
Kimi K3
Open Weight
Step 5 Preview
Open Weight
License
Kimi K3
Open Weight
Step 5 Preview
Open Weight
Release date
Kimi K3
2026-07-16
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
- Kimi K3 has the higher public point estimate, 69.46 versus 56.81. Their conditional score ranges overlap. These ranges do not establish rank confidence.
- Workload cost
- Repository review: $0.195 vs $0.0581. Cache-heavy agent loop: $0.27 vs $0.057.
- Context tradeoff
- Kimi K3 has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.
Questions
Which is better, Kimi K3 or Step 5 Preview?
Kimi K3 has the higher public point estimate, 69.46 versus 56.81. Their conditional score ranges 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, Kimi K3 or Step 5 Preview?
Kimi K3 scores higher for coding on the public lane, 58.6 to 48.8. 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, Kimi K3 or Step 5 Preview?
Kimi K3 scores higher for agentic tasks on the public lane, 68 to 55.3. Step 5 Preview 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, Kimi K3 or Step 5 Preview?
For the stated presets, chat costs $0.0105 on Kimi K3 and $0.00235 on Step 5 Preview; repository review costs $0.195 and $0.0581; the cache-heavy agent loop costs $0.27 and $0.057. Costs use the listed standard API rates.
Which has the larger context window, Kimi K3 or Step 5 Preview?
Kimi K3 has the larger documented context window: 1.05M, compared with 1M.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
Kimi K3
API / mo$13,500
Self-host / mo$57,606
Break-even291M/day
Step 5 Preview
API / mo$2,775
Self-host / mo$27,962
Break-even617M/day
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence65 rows
Agentic
Finch
Kimi K377.3%SourceStep 5 Preview—Not directly comparable
b3 Agent Security
Kimi K388.0%SourceStep 5 Preview—Not directly comparable
Cyber Misuse Refusal
Kimi K394.0%SourceStep 5 Preview—Not directly comparable
Terminal-Bench 2.1
Kimi K388.3%SourceStep 5 Preview85.0%SourceKimi K3 leads this result
BrowseComp
Kimi K391.2%SourceStep 5 Preview88.7%SourceKimi K3 leads this result
DeepSearchQA
Kimi K395.0%SourceStep 5 Preview—Not directly comparable
Toolathlon-Verified
Kimi K373.2%SourceStep 5 Preview74.1%SourceStep 5 Preview leads this result
MCP Atlas
Kimi K384.2%SourceStep 5 Preview85.6%SourceStep 5 Preview leads this result
AutomationBench
Kimi K330.8%SourceStep 5 Preview44.0%SourceStep 5 Preview leads this result
JobBench
Kimi K352.9%SourceStep 5 Preview59.0%SourceStep 5 Preview leads this result
APEX-Agents
Kimi K337.6%SourceStep 5 Preview37.8%SourceStep 5 Preview leads this result
SpreadsheetBench 2
Kimi K334.8%SourceStep 5 Preview—Not directly comparable
DECK-Bench
Kimi K373.5%SourceStep 5 Preview—Not directly comparable
Terminal-Bench 2.1 (Vals)
Kimi K380.9%SourceStep 5 Preview—Not directly comparable
ApprenticeBench
Kimi K318%SourceStep 5 Preview—Not directly comparable
Hermes Index
Shared sourceKimi K339.2%Step 5 Preview32.8%Kimi K3 leads this result
Hermes Bench
Shared sourceKimi K371.5%Step 5 Preview69.2%Kimi K3 leads this result
MazeBench
Kimi K31%SourceStep 5 Preview—Not directly comparable
Terminal-Bench 4.0
Kimi K3—Step 5 Preview33.30%SourceNot directly comparable
CyberGym
Kimi K3—Step 5 Preview84.7%SourceNot directly comparable
DRACO
Kimi K3—Step 5 Preview83.3%SourceNot directly comparable
Agents' Last Exam
Kimi K3—Step 5 Preview29.5%SourceNot directly comparable
Coding
ScarfBench
Kimi K361.3%SourceStep 5 Preview—Not directly comparable
DeepSWE
Kimi K367.5%SourceStep 5 Preview67.7%SourceStep 5 Preview leads this result
CursorBench 3.2
Kimi K360.8%SourceStep 5 Preview—Not directly comparable
FrontierSWE
Kimi K381.2%SourceStep 5 Preview—Not directly comparable
ProgramBench
Kimi K377.8%SourceStep 5 Preview80.5%SourceStep 5 Preview leads this result
Kimi Code Bench v2
Kimi K372.9%SourceStep 5 Preview—Not directly comparable
sweMarathon
Kimi K342%SourceStep 5 Preview72.7%SourceStep 5 Preview leads this result
PostTrain Bench
Kimi K336.6%SourceStep 5 Preview—Not directly comparable
MLS-Bench Lite
Kimi K348.3%SourceStep 5 Preview40.5%SourceKimi K3 leads this result
VulcanBench v3
Kimi K373.7%SourceStep 5 Preview—Not directly comparable
OpenHarmony Bench
Kimi K357.3%SourceStep 5 Preview—Not directly comparable
FrontierSWE v2
Kimi K325.9%SourceStep 5 Preview—Not directly comparable
LiveCodeBench (Vals)
Kimi K387.2%SourceStep 5 Preview—Not directly comparable
SWE-bench (Vals)
Kimi K393.4%SourceStep 5 Preview—Not directly comparable
PostTrainBench v1.1
Kimi K332.0%SourceStep 5 Preview—Not directly comparable
Bug Hunt Bench
Kimi K321.0 fixesSourceStep 5 Preview—Not directly comparable
FrontierCode 2.0
Kimi K317.7%SourceStep 5 Preview—Not directly comparable
Terminal-Bench 2.1
Kimi K3—Step 5 Preview85.0%SourceNot directly comparable
SciCode
Kimi K3—Step 5 Preview58.9%SourceNot directly comparable
Reasoning
Multimodal
ExtractBench (LlamaIndex)
Kimi K383.17%SourceStep 5 Preview—Not directly comparable
OfficeQA Pro
Kimi K363.3%SourceStep 5 Preview60.3%SourceKimi K3 leads this result
MMMU-Pro
Kimi K381.6%SourceStep 5 Preview76%SourceKimi K3 leads this result
MMMU-Pro w/ Python
Kimi K383.4%SourceStep 5 Preview—Not directly comparable
CharXiv w/o tools
Kimi K384.8%SourceStep 5 Preview—Not directly comparable
CharXiv
Kimi K391.3%SourceStep 5 Preview—Not directly comparable
MathVision
Kimi K394.3%SourceStep 5 Preview—Not directly comparable
MathVision w/ Python
Kimi K397.8%SourceStep 5 Preview—Not directly comparable
BabyVision w/ Python
Kimi K385.7%SourceStep 5 Preview—Not directly comparable
ZeroBench
Kimi K323.0%SourceStep 5 Preview—Not directly comparable
ZeroBench w/ Python
Kimi K341.0%SourceStep 5 Preview—Not directly comparable
WorldVQA ForceAnswer
Kimi K351.0%SourceStep 5 Preview—Not directly comparable
OmniDocBench
Kimi K391.1%SourceStep 5 Preview—Not directly comparable
PerceptionBench
Kimi K358.5%SourceStep 5 Preview—Not directly comparable
Knowledge
GPQA
Kimi K393.5%SourceStep 5 Preview—Not directly comparable
GPQA-D
Kimi K393.5%SourceStep 5 Preview93.5%SourceTie
HLE
Kimi K356%SourceStep 5 Preview46.5%SourceKimi K3 leads this result
HLE w/o tools
Kimi K343.5%SourceStep 5 Preview—Not directly comparable
GPQA Diamond (Vals)
Kimi K392.9%SourceStep 5 Preview—Not directly comparable
MMLU-Pro (Vals)
Kimi K388.0%SourceStep 5 Preview—Not directly comparable
Instruction following
Gray Swan IPI (15 attempts)
Kimi K352.7%SourceStep 5 Preview—Not directly comparable
65 public results · 17 shared
Watch Kimi K3 vs Step 5 Preview
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 6,100+ subscribers.
Last updated November 5, 2026