Multimodal
Like-for-like- GPT-5.4 nano
- 66.1
- Kimi K2.5
- 78.5
- Weighted basis
- 1 vs 1 rows
- Reading
- Kimi K2.5 leads
Model comparison
GPT-5.4 nano vs Kimi K2.5
Updated July 30, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
GPT-5.4 nano has the higher public score estimate, 65.98 versus 58.78, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
8 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
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.
Long documents
Prompts that approach the documented context limit
GPT-5.4 nano
GPT-5.4 nano has the larger documented context window.
Confidence: documented
Chat turn cost
1K fresh input + 500 output tokens
GPT-5.4 nano
GPT-5.4 nano 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
GPT-5.4 nano
GPT-5.4 nano has the lower estimated token cost for this stated workload. Kimi K2.5 has no published cached-input rate, so cached tokens use its listed input rate.
Confidence: rate-fallback
Show secondary and unsupported calls
Repository review cost
50K fresh input + 3K output tokens
GPT-5.4 nano
GPT-5.4 nano 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
No shared weighted benchmark basis supports a winner.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
5 GPT-5.4 nano only
8 Shared
37 Kimi K2.5 only
- Shared results
- 8
- GPT-5.4 nano only
- 5
- Kimi K2.5 only
- 37
- Like-for-like categories
- 1 / 8
3 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Category results, on a stated basis
Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.
Agentic
Directional only- GPT-5.4 nano
- 42.9
- Kimi K2.5
- 55.0
- Weighted basis
- 2 vs 2 rows
- Reading
- Directional only
Knowledge
Directional only- GPT-5.4 nano
- 43.8
- Kimi K2.5
- 56.9
- Weighted basis
- 2 vs 4 rows
- Reading
- Directional only
Math
Directional only- GPT-5.4 nano
- 21.0
- Kimi K2.5
- 60.6
- Weighted basis
- 2 vs 4 rows
- Reading
- Directional only
Coding
Not comparable- GPT-5.4 nano
- Not measured
- Kimi K2.5
- 59.4
- Weighted basis
- 0 vs 4 rows
- Reading
- Not comparable
Reasoning
Not comparable- GPT-5.4 nano
- Not measured
- Kimi K2.5
- 61.0
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Multilingual
Not comparable- GPT-5.4 nano
- Not measured
- Kimi K2.5
- 82.3
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Instruction following
Not comparable- GPT-5.4 nano
- Not measured
- Kimi K2.5
- 93.9
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
| Category | GPT-5.4 nano | Kimi K2.5 | Weighted basis | Reading |
|---|---|---|---|---|
| Multimodal | 66.1 | 78.5 | Like-for-like1 vs 1 rows | Kimi K2.5 leads |
| Agentic | 42.9 | 55.0 | Directional only2 vs 2 rows | Directional only |
| Knowledge | 43.8 | 56.9 | Directional only2 vs 4 rows | Directional only |
| Math | 21.0 | 60.6 | Directional only2 vs 4 rows | Directional only |
| Coding | Not measured | 59.4 | Not comparable0 vs 4 rows | Not comparable |
| Reasoning | Not measured | 61.0 | Not comparable0 vs 1 rows | Not comparable |
| Multilingual | Not measured | 82.3 | Not comparable0 vs 1 rows | Not comparable |
| Instruction following | Not measured | 93.9 | Not comparable0 vs 1 rows | Not comparable |
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.
MMMU-Pro
Multimodal
HLE
Knowledge
GPQA
Knowledge
Terminal-Bench 2.0
Agentic
FrontierMath v2 (Tier 4)
Math
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
- GPT-5.4 nano
- $0.00082
- Fits in one request
- Kimi K2.5
- $0.0021
- Fits in one request
GPT-5.4 nano has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- GPT-5.4 nano
- $0.01375
- Fits in one request
- Kimi K2.5
- $0.039
- Fits in one request
GPT-5.4 nano has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- GPT-5.4 nano
- $0.0205
- Fits in one request
- Kimi K2.5
- $0.162
- Fits in one request
- Cached input priced at the published list-input rate
GPT-5.4 nano has the lower modeled cost
Kimi K2.5 has no published cached-input rate, so cached tokens use its listed input rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationGPT-5.4 nanoKimi K2.5
Context window
Maximum documented context; output-token limits may be lower.
GPT-5.4 nano
Kimi K2.5
256K
API model ID
GPT-5.4 nano
gpt-5.4-nano
OpenAI GPT-5.4 nano model documentationKimi K2.5
Not sourced
Cached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
GPT-5.4 nano
$0.02 per 1M cached input tokens
OpenAI pricingKimi K2.5
Not published
Documented inputs
GPT-5.4 nano
text, image
OpenAI model catalogKimi K2.5
Not sourced
Documented outputs
GPT-5.4 nano
Kimi K2.5
Not sourced
Provider availability
GPT-5.4 nano
Generally Available · OpenAI Responses API
OpenAI model catalogKimi K2.5
Not sourced
Reasoning profile
GPT-5.4 nano
Reasoning
Kimi K2.5
Non-Reasoning
Weight access
GPT-5.4 nano
Proprietary
Kimi K2.5
Open Weight
License
GPT-5.4 nano
Proprietary
Kimi K2.5
Open Weight
Release date
GPT-5.4 nano
2026-03-17
Kimi K2.5
2026-02-01
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
- GPT-5.4 nano has the higher public score estimate, 65.98 versus 58.78, but the 90% score intervals overlap.
- Workload cost
- Repository review: $0.01375 vs $0.039. Cache-heavy agent loop: $0.0205 vs $0.162.
- Context tradeoff
- GPT-5.4 nano has the larger documented window (400K).
Run the same representative tasks against both endpoints before changing production traffic.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
GPT-5.4 nano
API / mo$1,088
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Kimi K2.5
API / mo$2,700
Self-host / mo$5,221
Break-even132M/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 evidence50 rows
Agentic
Terminal-Bench 2.0
GPT-5.4 nano46.3%SourceKimi K2.550.8%SourceKimi K2.5 leads this result
OSWorld-Verified
GPT-5.4 nano39%SourceKimi K2.5—Not directly comparable
MCP Atlas
GPT-5.4 nano56.1%SourceKimi K2.529.5%SourceGPT-5.4 nano leads this result
Toolathlon
GPT-5.4 nano35.5%SourceKimi K2.527.8%SourceGPT-5.4 nano leads this result
τ²-bench results
GPT-5.4 nano92.5%SourceKimi K2.5—Not directly comparable
BrowseComp
GPT-5.4 nano—Kimi K2.560.6%SourceNot directly comparable
Claw-Eval
GPT-5.4 nano—Kimi K2.552.3%SourceNot directly comparable
QwenClawBench
GPT-5.4 nano—Kimi K2.554.3%SourceNot directly comparable
τ³-bench results
GPT-5.4 nano—Kimi K2.565.7%SourceNot directly comparable
DeepSearchQA
GPT-5.4 nano—Kimi K2.577.1%SourceNot directly comparable
DeepPlanning
GPT-5.4 nano—Kimi K2.514.4%SourceNot directly comparable
MCP-Tasks
GPT-5.4 nano—Kimi K2.559.1%SourceNot directly comparable
WideResearch
GPT-5.4 nano—Kimi K2.572.7%SourceNot directly comparable
Gert Labs
GPT-5.4 nano—Kimi K2.545.88%SourceNot directly comparable
ResearchClawBench
GPT-5.4 nano—Kimi K2.514.0%SourceNot directly comparable
JobBench
GPT-5.4 nano—Kimi K2.58.7%SourceNot directly comparable
Coding
Vibe Code Bench
GPT-5.4 nano26.10%SourceKimi K2.5—Not directly comparable
SWE-bench Verified
GPT-5.4 nano—Kimi K2.576.8%SourceNot directly comparable
SWE-bench Verified*
GPT-5.4 nano—Kimi K2.570.8%SourceNot directly comparable
LiveCodeBench v6
GPT-5.4 nano—Kimi K2.585.0%SourceNot directly comparable
SWE-bench Pro
GPT-5.4 nano—Kimi K2.550.7%SourceNot directly comparable
SWE Multilingual
GPT-5.4 nano—Kimi K2.573%SourceNot directly comparable
SWE-Rebench
GPT-5.4 nano—Kimi K2.558.5%SourceNot directly comparable
React Native Evals
GPT-5.4 nano—Kimi K2.577.2%SourceNot directly comparable
SciCode
GPT-5.4 nano—Kimi K2.548.7%SourceNot directly comparable
Reasoning
LongBench v2
GPT-5.4 nano—Kimi K2.561%SourceNot directly comparable
Knowledge
GPQA
GPT-5.4 nano82.8%SourceKimi K2.587.6%SourceKimi K2.5 leads this result
HLE
GPT-5.4 nano37.7%SourceKimi K2.530.1%SourceGPT-5.4 nano leads this result
HLE w/o tools
GPT-5.4 nano24.3%SourceKimi K2.5—Not directly comparable
GPQA-D
GPT-5.4 nano—Kimi K2.587.6%SourceNot directly comparable
SuperGPQA
GPT-5.4 nano—Kimi K2.569.2%SourceNot directly comparable
MMLU-Pro
GPT-5.4 nano—Kimi K2.587.1%SourceNot directly comparable
MMLU-Pro (Arcee)
GPT-5.4 nano—Kimi K2.587.1%SourceNot directly comparable
Math
FrontierMath v2 (Tiers 1-3)
Shared sourceGPT-5.4 nano25.860%Kimi K2.527.900%Kimi K2.5 leads this result
FrontierMath v2 (Tier 4)
Shared sourceGPT-5.4 nano6.250%Kimi K2.54.200%GPT-5.4 nano leads this result
AIME 2025
GPT-5.4 nano—Kimi K2.596.1%SourceNot directly comparable
AIME26
GPT-5.4 nano—Kimi K2.595.8%SourceNot directly comparable
AIME25 (Arcee)
GPT-5.4 nano—Kimi K2.596.3%SourceNot directly comparable
HMMT Feb 2025
GPT-5.4 nano—Kimi K2.595.4%SourceNot directly comparable
HMMT Nov 2025
GPT-5.4 nano—Kimi K2.591.1%SourceNot directly comparable
HMMT Feb 2026
GPT-5.4 nano—Kimi K2.587.1%SourceNot directly comparable
MMAnswerBench
GPT-5.4 nano—Kimi K2.581.8%SourceNot directly comparable
Multilingual
Multimodal
MMMU-Pro
GPT-5.4 nano66.1%SourceKimi K2.578.5%SourceKimi K2.5 leads this result
MMMU-Pro w/ Python
GPT-5.4 nano69.5%SourceKimi K2.5—Not directly comparable
Video-MME
GPT-5.4 nano—Kimi K2.587.4%SourceNot directly comparable
MMVU
GPT-5.4 nano—Kimi K2.580.4%SourceNot directly comparable
VideoMMMU
GPT-5.4 nano—Kimi K2.586.6%SourceNot directly comparable
Instruction following
IFEval
GPT-5.4 nano—Kimi K2.593.9%SourceNot directly comparable
Frequently asked questions
Which is better, GPT-5.4 nano or Kimi K2.5?
GPT-5.4 nano has the higher public score estimate, 65.98 versus 58.78, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Which is better for coding, GPT-5.4 nano or Kimi K2.5?
The published evidence does not provide a shared weighted coding basis for both models, so BenchLM does not name a coding winner.
Which is better for agentic tasks, GPT-5.4 nano or Kimi K2.5?
The current agentic tasks averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.
Which costs less, GPT-5.4 nano or Kimi K2.5?
For the stated presets, chat costs $0.00082 on GPT-5.4 nano and $0.0021 on Kimi K2.5; repository review costs $0.01375 and $0.039; the cache-heavy agent loop costs $0.0205 and $0.162. Kimi K2.5 has no published cached-input rate, so cached tokens use its listed input rate.
Which has the larger context window, GPT-5.4 nano or Kimi K2.5?
GPT-5.4 nano has the larger documented context window: 400K, compared with 256K.
Related comparisons
Last updated July 30, 2026
Watch GPT-5.4 nano vs Kimi K2.5
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.