Coding
Like-for-like- Hy3 Preview
- 74.4
- Kimi K2.5 (Reasoning)
- 76.8
- Weighted basis
- 1 vs 1 rows
- Reading
- Kimi K2.5 (Reasoning) leads
Provider changes are easy to miss. Radar watches releases, pricing, deprecations, and incidents at the source.Provider changes are easy to miss.
See RadarModel comparison
Updated August 7, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.
Kimi K2.5 (Reasoning) has the higher public score estimate, 58.53 versus 42.8, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.
4 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.
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.
Code generation, repair, and software-engineering tasks
Kimi K2.5 (Reasoning)
Kimi K2.5 (Reasoning) leads on the same 1 weighted benchmark row.
Confidence: limited
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
Prompts that approach the documented context limit
No clear pick
The documented context windows are equal.
Confidence: documented
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
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
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
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
2 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.
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.
| Category | Hy3 Preview | Kimi K2.5 (Reasoning) | Weighted basis | Reading |
|---|---|---|---|---|
| Coding | 74.4 | 76.8 | Like-for-like1 vs 1 rows | Kimi K2.5 (Reasoning) leads |
| Agentic | 54.4 | 55.0 | Directional only1 vs 2 rows | Directional only |
| Knowledge | 87.2 | 87.2 | Directional only1 vs 2 rows | Directional only |
| Reasoning | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Math | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Multilingual | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
| Multimodal | Not measured | 78.5 | Not comparable0 vs 1 rows | Not comparable |
| Instruction following | Not measured | Not measured | Not comparable0 vs 0 rows | Not comparable |
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.
Terminal-Bench 2.0
Agentic
SWE-bench Verified
Coding
GPQA
Knowledge
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.
1K fresh input + 500 output tokens
Hy3 Preview has no comparable published API token rate.
50K fresh input + 3K output tokens
Hy3 Preview has no comparable published API token rate.
200K cached + 20K fresh input + 10K output tokens
Kimi K2.5 (Reasoning) has no published cached-input rate, so cached tokens use its listed input rate. Hy3 Preview has no comparable published API token rate.
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
Maximum documented context; output-token limits may be lower.
Hy3 Preview
256K
Kimi K2.5 (Reasoning)
256K
Hy3 Preview
Not sourced
Kimi K2.5 (Reasoning)
Not sourced
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
Hy3 Preview
No comparable hosted API rate
Kimi K2.5 (Reasoning)
Not published
Hy3 Preview
Not sourced
Kimi K2.5 (Reasoning)
Not sourced
Hy3 Preview
Not sourced
Kimi K2.5 (Reasoning)
Not sourced
Hy3 Preview
Not sourced
Kimi K2.5 (Reasoning)
Not sourced
Hy3 Preview
Reasoning
Kimi K2.5 (Reasoning)
Reasoning
Hy3 Preview
Open Weight
Kimi K2.5 (Reasoning)
Proprietary
Hy3 Preview
Open Weight
Kimi K2.5 (Reasoning)
Proprietary
Hy3 Preview
2026-04-23
Kimi K2.5 (Reasoning)
2026-02-01
Run the same representative tasks against both endpoints before changing production traffic.
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Terminal-Bench 2.0
Hy3 Preview leads this result
Gert Labs
Shared sourceHy3 Preview leads this result
BrowseComp
Not directly comparable
AIME 2025
Not directly comparable
MMMU-Pro
Not directly comparable
Kimi K2.5 (Reasoning) has the higher public score estimate, 58.53 versus 42.8, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
Kimi K2.5 (Reasoning) leads the like-for-like coding comparison across 1 shared weighted benchmark row.
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.
Both models do not have comparable published API token rates, so this page does not name a universal price winner.
Both models list the same context window, 256K.
Last updated August 7, 2026
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.