Across 19 comparable benchmarks, Kimi K2.6 has better reported results on 17 and worse on 2. 1 further shared result is listed below, each with the reason it is not counted.
Moderate evidence19 of 20 shared results countedHow we compare
Price basis: Standard rates vs international rates. The price section below shows the applicable periods.
| Fact | GPT-5-mini | Kimi K2.6 |
|---|---|---|
| Context window | 400K tokens | 256K tokens |
| Input / 1M | $0.250 | $0.950 |
| Output / 1M | $2.00 | $4.00 |
| Blended (4:1) | $0.438 | $0.991 |
| Status | Deprecated | Available |
| Released | August 7, 2025 | April 20, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 11 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Kimi K2.6 ahead on 4 of 4 counted
Kimi K2.6 ahead on 2 of 2 counted
Kimi K2.6 ahead on 3 of 5 counted
Kimi K2.6 ahead on 2 of 2 counted
Kimi K2.6 ahead on 3 of 3 counted
No shared comparable results for Abstract reasoning, Computer & tool use, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show GPT-5-mini relative to Kimi K2.6. Reading the results
| Benchmark | GPT-5-mini | Kimi K2.6 | Gap |
|---|---|---|---|
| LiveCodeBenchAlgorithmic Coding · Pass @K | 55.0Thinking · no tools | 89.6Thinking · no tools | 0.61×#80 vs #10 of 113 |
| SWE-bench MultilingualRepository Engineering · Accuracy | 39.7Unspecified Reasoning · tools | 76.7≠Thinking · tools | −37.0 pts#48 vs #17 of 48 |
| Terminal Bench HardAgentic Development · Accuracy | 33.3Thinking High · tools | 43.9≠Thinking · tools | −10.6 pts#58 vs #23 of 157 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 14.2Thinking High · tools | 37.9≠Unspecified Reasoning · tools | −23.7 pts#97 vs #74 of 121 |
| Vending-Bench 2Logical Planning · Money Balance (USD)Not counted · Speed and cost telemetry is reported, not scored | -31.2Unspecified Reasoning · tools | 6205Unspecified Reasoning · tools | —#64 vs #27 of 64 |
| FrontierMath Tier 4 v2Mathematics · Accuracy (verification_code) | 12.2Thinking High · tools | 25.6≠Reported best | −13.4 pts#59 vs #46 of 73 |
| FrontierMath v2Mathematics · Accuracy (verification_code) | 46.7Thinking High · tools | 57.2≠Reported best | −10.5 pts#57 vs #48 of 85 |
| CritPtScientific Reasoning · Score | 1.4Thinking Medium · no tools | 8.0≠Thinking · no tools | 0.17×#96 vs #62 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 69.0Thinking · no tools | 90.5Thinking · no tools | −21.5 pts#146 vs #46 of 200 |
| SciCodeScientific Computing · Score | 39.0Thinking High · no tools | 51.5≠Thinking · no tools | 0.76×#77 vs #45 of 98 |
| MedCodeClinical Workflows · Accuracy (%) | 43.0Thinking High · tools | 40.1≠Unspecified Reasoning · tools | +2.9 pts#46 vs #63 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 80.6Thinking High · tools | 78.2≠Unspecified Reasoning · tools | +2.4 pts#52 vs #61 of 99 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 8.4Thinking High · no tools | 13.0≠Thinking · no tools | −4.6 pts#64 vs #49 of 96 |
| MMMU-ProVisual Understanding · Accuracy | 70.1Thinking High · no tools | 79.4≠Thinking · no tools | −9.3 pts#83 vs #36 of 136 |
| Creative WritingWriting | 1314Normal · no tools | 1721Normal · no tools | 0.76×#84 vs #30 of 121 |
| SAGEService Workflows · Accuracy (%) | 43.0Thinking High · tools | 50.2≠Unspecified Reasoning · tools | −7.2 pts#52 vs #17 of 82 |
| τ³-BankingService Workflows · Score | 15.5Thinking High · tools | 23.3≠Thinking · tools | 0.67×#71 vs #51 of 120 |
| HLEKnowledge Exams · Accuracy | 19.4Reported best | 54.0≠Thinking · tools + web | −34.6 pts#105 vs #33 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 146Reported best | 151Reported best | 0.96×#77 vs #45 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 65.9Thinking High · no tools | 70.5≠Thinking · no tools | −4.6 pts#70 vs #56 of 106 |
Algorithmic Coding · Pass @K
Repository Engineering · Accuracy
Agentic Development · Accuracy
Code Generation & Editing · Pass rate (%)
Logical Planning · Money Balance (USD)
Not counted · Speed and cost telemetry is reported, not scored
Mathematics · Accuracy (verification_code)
Mathematics · Accuracy (verification_code)
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Scientific Computing · Score
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Writing
Service Workflows · Accuracy (%)
Service Workflows · Score
A dash means no result is available.
| Benchmark | GPT-5-mini | Kimi K2.6 |
|---|---|---|
| AIME 2026 | — | 96.4 |
| AIME2025 | 87.5 | — |
| FrontierMath | 19.3 | — |
| FrontierMath - Tier 4 | 6.3 | — |
| IMO-AnswerBench | — | 86.0 |
| MathArena Apex | 1.0 | — |
| MMLU-Pro | 78.0 | — |
| Riemann-bench | — | 7.2 |
| Code Migration | — | 27.8 |
| CodeClash | 1200 | — |
| FrontierSWE v1 | — | 22.2 |
| Kimi Code Bench 2.0 | — | 50.9 |
| MLS Bench | — | 26.7 |
| SWE-Bench Pro - Public | — | 58.6 |
| SWE-bench Verified | — | 80.2 |
| Terminal Bench 2.0 | — | 66.7 |
| Terminal-Bench 1.0 | 14.0 | — |
| Terminal-Bench 2.1 | — | 53.6 |
| WeirdML v2 | 52.7 | — |
| Agents' Last Exam v1.0 | — | 9.2 |
| Claw Bench | — | 80.9 |
| MCP-Atlas | — | 69.4 |
| MCPMark-Verified | — | 72.8 |
| OSWorld 2.0 | — | 22.1 |
| OSWorld-Verified | — | 73.1 |
| Pinch Bench | 80.3 | — |
| Tool Decathlon | — | 50.0 |
| Toolathlon-Verified | — | 58.0 |
| ExploitBench | — | 18.4 |
| Context Arena | — | 64.6 |
| EBR-bench | — | 2.4 |
| Fiction.liveBench | 69.4 | — |
| VPCT | 40.2 | — |
| AutomationBench | — | 2.6 |
| EMB (Excel Modeling) | — | 57.9 |
| Finance Agent v2 | — | 44.9 |
| Harvey Lab-AA | — | 84.1 |
| Legal Research Bench | — | 15.9 |
| Public Benefits Bench v1.1 | — | 56.6 |
| Tax Agent Bench | — | 12.4 |
| ARC-AGI-1 | 54.3 | — |
| ARC-AGI-2 | 4.4 | — |
| BrowseComp | — | 83.2 |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
GPT-5-mini costs56%lessthan Kimi K2.6 per million tokens — 2.3× cheaper.Using standard rates vs international rates at this token mix.
GPT-5-mini: Standard pricing. OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-11-13; standard tier; Deprecated; shutdown 2026-12-11 (replacement: gpt-6.1-terra)
Kimi K2.6: International pricing. Kimi Intl standard
| Price / 1M tokens | GPT-5-mini | Kimi K2.6 | Ratio |
|---|---|---|---|
| Input | $0.25 | $0.95 | 0.26× |
| Output | $2.00 | $4.00 | 0.50× |
| Cached input | $0.025 | $0.16 | 0.16× |
| Cache write | — | — | — |
| Blended | $0.438 | $0.991 | 0.44× |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
| Benchmark | GPT-5-mini | Kimi K2.6 |
|---|---|---|
| Agents' Last Exam v1.0 | Not reported | |
| AIME 2026 | Not reported | |
| ARC-AGI-1 | Not reported | |
| ARC-AGI-2 | $0.1977 / task Average · ARC-AGI-2 Score: 4.4 Thinking High (With Tools) $0.0189 / task Average · ARC-AGI-2 Score: 0.83 Thinking Low (With Tools) $0.0629 / task Average · ARC-AGI-2 Score: 4.0 Thinking Medium (With Tools) | Not reported |
| AutomationBench | Not reported | $0.28 / task Average · Pass Rate: 2.6 Reported best (effort unspecified) · zapier-private-heldout |
| Code Migration | Not reported | $5.1121 / task Average · Hidden-test pass rate (%): 27.8 Unspecified Reasoning (With Tools) |
| EMB (Excel Modeling) | Not reported | |
| Finance Agent v2 | Not reported | |
| Legal Research Bench | Not reported | |
| MedCode | Not reported | |
| MedScribe | Not reported | |
| Public Benefits Bench v1.1 | Not reported | $0.9063 / task Average · Criterion pass rate (%): 56.6 Unspecified Reasoning (With Tools) |
| SAGE | Not reported | |
| Tax Agent Bench | Not reported | $0.5655 / task Average · Accuracy (%, rubric × citation quality): 12.4 Unspecified Reasoning (With Tools) |
| Vibe Code Bench v1.1 | Not reported |
| Attribute | GPT-5-mini | Kimi K2.6 |
|---|---|---|
| Availability | Deprecated | Available |
| Provider | OpenAI | Moonshot AI |
| Model type | Foundation model | Reasoning model |
| First release | August 7, 2025 | April 20, 2026 |
| Context window | 400K tokens | 256K tokens |
| Max output | 128K tokens | — |
| Total parameters | — | 1.0T |
| Active parameters | — | 32B |
| Architecture | Undisclosed | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual, Coding, Tool use, Function calling, Computer use, Long context |
| Knowledge cutoff | — | 2025-04 |
| Input modalities | Text, Image | Text, Image, Video |
| Output modalities | Text | Text |
| Weights | Closed / API only | Open weights |
| Licence | — | Modified MIT |
| Commercial use | Proprietary | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used GPT-5-mini or Kimi K2.6? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.