Coding
Directional only- Hy4 preview
- 65.7
- Kimi K2.5
- 59.4
- Weighted basis
- 1 vs 4 rows
- Reading
- Directional only
Hy4 preview vs Kimi K2.5
Updated August 28, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Decision reading
Hy4 preview has the higher public score, 79.16 versus 59.14, and the 90% score intervals do not overlap.
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
Hy4 preview
Hy4 preview has the larger documented context window.
Confidence: documented
Show secondary and unsupported calls
Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
No shared weighted benchmark basis supports a winner.
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: rate-fallback
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
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
20 Hy4 preview only
8 Shared
37 Kimi K2.5 only
- Shared results
- 8
- Hy4 preview only
- 20
- Kimi K2.5 only
- 37
- Like-for-like categories
- 0 / 8
2 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.
Knowledge
Directional only- Hy4 preview
- 60.4
- Kimi K2.5
- 56.9
- Weighted basis
- 2 vs 4 rows
- Reading
- Directional only
Agentic
Not comparable- Hy4 preview
- Not measured
- Kimi K2.5
- 55.0
- Weighted basis
- 0 vs 2 rows
- Reading
- Not comparable
Reasoning
Not comparable- Hy4 preview
- Not measured
- Kimi K2.5
- 61.0
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Math
Not comparable- Hy4 preview
- Not measured
- Kimi K2.5
- 60.6
- Weighted basis
- 0 vs 4 rows
- Reading
- Not comparable
Multilingual
Not comparable- Hy4 preview
- Not measured
- Kimi K2.5
- 82.3
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
Multimodal
Not comparable- Hy4 preview
- 66.2
- Kimi K2.5
- 78.5
- Weighted basis
- 1 vs 1 rows
- Reading
- Not comparable
Instruction following
Not comparable- Hy4 preview
- Not measured
- Kimi K2.5
- 93.9
- Weighted basis
- 0 vs 1 rows
- Reading
- Not comparable
| Category | Hy4 preview | Kimi K2.5 | Weighted basis | Reading |
|---|---|---|---|---|
| Coding | 65.7 | 59.4 | Directional only1 vs 4 rows | Directional only |
| Knowledge | 60.4 | 56.9 | Directional only2 vs 4 rows | Directional only |
| Agentic | Not measured | 55.0 | Not comparable0 vs 2 rows | Not comparable |
| Reasoning | Not measured | 61.0 | Not comparable0 vs 1 rows | Not comparable |
| Math | Not measured | 60.6 | Not comparable0 vs 4 rows | Not comparable |
| Multilingual | Not measured | 82.3 | Not comparable0 vs 1 rows | Not comparable |
| Multimodal | 66.2 | 78.5 | Not comparable1 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.
HLE
Knowledge
SWE-bench Pro
Coding
GPQA
Knowledge
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
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Kimi K2.5
- $0.0021
- Fits in one request
Hy4 preview has no comparable published API token rate.
Repository review
50K fresh input + 3K output tokens
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Kimi K2.5
- $0.039
- Fits in one request
Hy4 preview has no comparable published API token rate.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- Hy4 preview
- Self-hosted; infrastructure cost varies
- Fits in one request
- Cached-input rate unavailable
- Kimi K2.5
- $0.162
- Fits in one request
- Cached input priced at the published list-input rate
Kimi K2.5 has no published cached-input rate, so cached tokens use its listed input rate. Hy4 preview has no comparable published API token rate.
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationHy4 previewKimi K2.5
Context window
Maximum documented context; output-token limits may be lower.
Hy4 preview
Kimi K2.5
256K
API model ID
Hy4 preview
tencent/Hy4-preview
Tencent Hy4 preview model cardKimi 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.
Hy4 preview
No comparable hosted API rate
Tencent Hy4 preview model cardKimi K2.5
Not published
Documented inputs
Hy4 preview
Not sourced
Kimi K2.5
Not sourced
Documented outputs
Hy4 preview
Not sourced
Kimi K2.5
Not sourced
Provider availability
Hy4 preview
Not sourced
Kimi K2.5
Not sourced
Reasoning profile
Hy4 preview
Reasoning
Kimi K2.5
Non-Reasoning
Weight access
Hy4 preview
Open Weight
Kimi K2.5
Open Weight
License
Hy4 preview
Open Weight
Kimi K2.5
Open Weight
Release date
Hy4 preview
2026-08-28
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
- Hy4 preview has the higher public score, 79.16 versus 59.14, and the 90% score intervals do not overlap.
- Workload cost
- A complete comparable API-rate estimate is not available for both models.
- Context tradeoff
- Hy4 preview has the larger documented window (1M).
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.
Hy4 preview
API / mo$0
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 evidence65 rows
Agentic
Terminal-Bench 2.1
Hy4 preview85.4%SourceKimi K2.5—Not directly comparable
CyberGym
Hy4 preview78.4%SourceKimi K2.5—Not directly comparable
WideResearch
Hy4 preview83.9%SourceKimi K2.572.7%SourceHy4 preview leads this result
DRACO
Hy4 preview77.2%SourceKimi K2.5—Not directly comparable
MCP Atlas
Hy4 preview83.7%SourceKimi K2.529.5%SourceHy4 preview leads this result
Toolathlon-Verified
Hy4 preview74.1%SourceKimi K2.5—Not directly comparable
APEX-Agents
Hy4 preview37.1%SourceKimi K2.5—Not directly comparable
skillsBench
Hy4 preview62.9%SourceKimi K2.5—Not directly comparable
JobBench
Hy4 preview61.7%SourceKimi K2.58.7%SourceHy4 preview leads this result
Agents' Last Exam
Hy4 preview22.8%SourceKimi K2.5—Not directly comparable
AutomationBench
Hy4 preview32.1%SourceKimi K2.5—Not directly comparable
BankerToolBench
Hy4 preview78.6%SourceKimi K2.5—Not directly comparable
HLE w/ tools
Hy4 preview55.4%SourceKimi K2.5—Not directly comparable
Terminal-Bench 2.0
Hy4 preview—Kimi K2.550.8%SourceNot directly comparable
BrowseComp
Hy4 preview—Kimi K2.560.6%SourceNot directly comparable
Claw-Eval
Hy4 preview—Kimi K2.552.3%SourceNot directly comparable
QwenClawBench
Hy4 preview—Kimi K2.554.3%SourceNot directly comparable
τ³-bench results
Hy4 preview—Kimi K2.565.7%SourceNot directly comparable
DeepSearchQA
Hy4 preview—Kimi K2.577.1%SourceNot directly comparable
DeepPlanning
Hy4 preview—Kimi K2.514.4%SourceNot directly comparable
Toolathlon
Hy4 preview—Kimi K2.527.8%SourceNot directly comparable
MCP-Tasks
Hy4 preview—Kimi K2.559.1%SourceNot directly comparable
Gert Labs
Hy4 preview—Kimi K2.545.88%SourceNot directly comparable
ResearchClawBench
Hy4 preview—Kimi K2.514.0%SourceNot directly comparable
Coding
Terminal-Bench 2.1
Hy4 preview85.4%SourceKimi K2.5—Not directly comparable
SWE-bench Pro
Hy4 preview65.7%SourceKimi K2.550.7%SourceHy4 preview leads this result
SWE Multilingual
Hy4 preview82.9%SourceKimi K2.573%SourceHy4 preview leads this result
deepSwe
Hy4 preview64.3%SourceKimi K2.5—Not directly comparable
NL2Repo
Hy4 preview58.9%SourceKimi K2.5—Not directly comparable
ProgramBench
Hy4 preview17.5%SourceKimi K2.5—Not directly comparable
PostTrain Bench
Hy4 preview35.6%SourceKimi K2.5—Not directly comparable
sweMarathon
Hy4 preview31.9%SourceKimi K2.5—Not directly comparable
SWE-bench Verified
Hy4 preview—Kimi K2.576.8%SourceNot directly comparable
SWE-bench Verified*
Hy4 preview—Kimi K2.570.8%SourceNot directly comparable
LiveCodeBench v6
Hy4 preview—Kimi K2.585.0%SourceNot directly comparable
SWE-Rebench
Hy4 preview—Kimi K2.558.5%SourceNot directly comparable
React Native Evals
Hy4 preview—Kimi K2.577.2%SourceNot directly comparable
SciCode
Hy4 preview—Kimi K2.548.7%SourceNot directly comparable
Reasoning
Knowledge
GPQA
Hy4 preview92.3%SourceKimi K2.587.6%SourceHy4 preview leads this result
GPQA-D
Hy4 preview92.3%SourceKimi K2.587.6%SourceHy4 preview leads this result
HLE
Hy4 preview55.4%SourceKimi K2.530.1%SourceHy4 preview leads this result
HLE w/o tools
Hy4 preview43.4%SourceKimi K2.5—Not directly comparable
SuperGPQA
Hy4 preview—Kimi K2.569.2%SourceNot directly comparable
MMLU-Pro
Hy4 preview—Kimi K2.587.1%SourceNot directly comparable
MMLU-Pro (Arcee)
Hy4 preview—Kimi K2.587.1%SourceNot directly comparable
Math
Apex
Hy4 preview74.2%SourceKimi K2.5—Not directly comparable
AIME 2025
Hy4 preview—Kimi K2.596.1%SourceNot directly comparable
AIME26
Hy4 preview—Kimi K2.595.8%SourceNot directly comparable
AIME25 (Arcee)
Hy4 preview—Kimi K2.596.3%SourceNot directly comparable
HMMT Feb 2025
Hy4 preview—Kimi K2.595.4%SourceNot directly comparable
HMMT Nov 2025
Hy4 preview—Kimi K2.591.1%SourceNot directly comparable
HMMT Feb 2026
Hy4 preview—Kimi K2.587.1%SourceNot directly comparable
MMAnswerBench
Hy4 preview—Kimi K2.581.8%SourceNot directly comparable
FrontierMath v2 (Tiers 1-3)
Hy4 preview—Kimi K2.527.900%SourceNot directly comparable
FrontierMath v2 (Tier 4)
Hy4 preview—Kimi K2.54.200%SourceNot directly comparable
Multilingual
Multimodal
OfficeQA Pro
Hy4 preview66.2%SourceKimi K2.5—Not directly comparable
MMMU-Pro
Hy4 preview—Kimi K2.578.5%SourceNot directly comparable
Video-MME
Hy4 preview—Kimi K2.587.4%SourceNot directly comparable
MMVU
Hy4 preview—Kimi K2.580.4%SourceNot directly comparable
VideoMMMU
Hy4 preview—Kimi K2.586.6%SourceNot directly comparable
Instruction following
IFEval
Hy4 preview—Kimi K2.593.9%SourceNot directly comparable
Frequently asked questions
Which is better, Hy4 preview or Kimi K2.5?
Hy4 preview has the higher public score, 79.16 versus 59.14, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.
Which is better for coding, Hy4 preview or Kimi K2.5?
The current coding 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 is better for agentic tasks, Hy4 preview or Kimi K2.5?
The published evidence does not provide a shared weighted agentic tasks basis for both models, so BenchLM does not name a agentic tasks winner.
Which costs less, Hy4 preview or Kimi K2.5?
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, Hy4 preview or Kimi K2.5?
Hy4 preview has the larger documented context window: 1M, compared with 256K.
Related comparisons
Last updated August 28, 2026
Watch Hy4 preview 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.