Across 4 comparable benchmarks, Kimi K2.6 has better reported results on 4 and worse on 0.
Moderate evidence4 of 4 shared results countedHow we compare
GPT-5 Codex is no longer available through its official API. API prices for this version are not shown.
Price basis: International rates vs listed rates. The price section below shows the applicable periods.
| Fact | Kimi K2.6 | GPT-5 Codex |
|---|---|---|
| Context window | 256K tokens | 400K tokens |
| Input / 1M | $0.950 | — |
| Output / 1M | $4.00 | — |
| Blended (4:1) | $0.991 | — |
| Status | Available | Retired |
| Released | April 20, 2026 | September 15, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 3 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Kimi K2.6 ahead on 2 of 2 counted
Kimi K2.6 ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
Kimi K2.6 ahead on 1 of 1 counted
No shared comparable results for Abstract reasoning, Mathematics, Computer & tool use, Research & search, Long context, Professional work & productivity, Cybersecurity.
Published scores, with matching modes where available. Gaps show Kimi K2.6 relative to GPT-5 Codex. Reading the results
| Benchmark | Kimi K2.6 | GPT-5 Codex | Gap |
|---|---|---|---|
| SWE-bench VerifiedRepository Engineering · Accuracy | 80.2Thinking · tools | 74.5≠Thinking High · no tools | +5.7 pts#15 vs #39 of 108 |
| Terminal Bench HardAgentic Development · Accuracy | 43.9Thinking · tools | 37.9≠Thinking High · tools | +6.0 pts#23 vs #38 of 157 |
| CritPtScientific Reasoning · Score | 8.0Thinking · no tools | 5.1≠Thinking High · no tools | 1.57×#62 vs #70 of 134 |
| MMMU-ProVisual Understanding · Accuracy | 79.4Thinking · no tools | 73.8≠Thinking High · no tools | +5.6 pts#36 vs #74 of 136 |
Repository Engineering · Accuracy
Agentic Development · Accuracy
Scientific Reasoning · Score
Visual Understanding · Accuracy
A dash means no result is available.
| Benchmark | Kimi K2.6 | GPT-5 Codex |
|---|---|---|
| AIME 2026 | 96.4 | — |
| AIME2025 | — | 98.7 |
| FrontierMath Tier 4 v2 | 25.6 | — |
| FrontierMath v2 | 57.2 | — |
| GPQA Diamond | 90.5 | — |
| HLE | 54.0 | — |
| IMO-AnswerBench | 86.0 | — |
| MMLU-Pro | — | 86.5 |
| Riemann-bench | 7.2 | — |
| SciCode | 51.5 | — |
| Code Migration | 27.8 | — |
| FrontierSWE v1 | 22.2 | — |
| Kimi Code Bench 2.0 | 50.9 | — |
| LiveCodeBench | 89.6 | — |
| MLS Bench | 26.7 | — |
| SWE-bench Multilingual | 76.7 | — |
| SWE-Bench Pro - Public | 58.6 | — |
| Terminal Bench 2.0 | 66.7 | — |
| Terminal-Bench 2.1 | 53.6 | — |
| Vibe Code Bench v1.1 | 37.9 | — |
| 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 | — |
| Tool Decathlon | 50.0 | — |
| Toolathlon-Verified | 58.0 | — |
| ExploitBench | 18.4 | — |
| ECI | 151 | — |
| LiveBench | 70.5 | — |
| Creative Writing | 1721 | — |
| Context Arena | 64.6 | — |
| EBR-bench | 2.4 | — |
| GDP.pdf | 13.0 | — |
| 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 | — |
| SAGE | 50.2 | — |
| Tax Agent Bench | 12.4 | — |
| τ³-Banking | 23.3 | — |
| Vending-Bench 2 | 6205 | — |
| BrowseComp | 83.2 | — |
| MedCode | 40.1 | — |
| MedScribe | 78.2 | — |
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%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
Kimi K2.6: International pricing. Kimi Intl standard
GPT-5 Codex: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | Kimi K2.6 | GPT-5 Codex | Ratio |
|---|---|---|---|
| Input | $0.95 | — | — |
| Output | $4.00 | — | — |
| Cached input | $0.16 | — | — |
| Cache write | — | — | — |
| Blended | $0.991 | — | — |
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 | Kimi K2.6 | GPT-5 Codex |
|---|---|---|
| Agents' Last Exam v1.0 | Not reported | |
| AIME 2026 | Not reported | |
| AutomationBench | $0.28 / task Average · Pass Rate: 2.6 Reported best (effort unspecified) · zapier-private-heldout | Not reported |
| Code Migration | $5.1121 / task Average · Hidden-test pass rate (%): 27.8 Unspecified Reasoning (With Tools) | Not reported |
| 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 | $0.9063 / task Average · Criterion pass rate (%): 56.6 Unspecified Reasoning (With Tools) | Not reported |
| SAGE | Not reported | |
| Tax Agent Bench | $0.5655 / task Average · Accuracy (%, rubric × citation quality): 12.4 Unspecified Reasoning (With Tools) | Not reported |
| Vibe Code Bench v1.1 | Not reported |
| Attribute | Kimi K2.6 | GPT-5 Codex |
|---|---|---|
| Availability | Available | Retired |
| Provider | Moonshot AI | OpenAI |
| Model type | Reasoning model | Coding model |
| First release | April 20, 2026 | September 15, 2025 |
| Context window | 256K tokens | 400K tokens |
| Max output | — | 128K tokens |
| Total parameters | 1.0T | — |
| Active parameters | 32B | — |
| Architecture | Mixture of Experts (MoE) | Undisclosed |
| Documented capabilities | Reasoning, Multilingual, Coding, Tool use, Function calling, Computer use, Long context | Reasoning, Multilingual, Coding |
| Knowledge cutoff | 2025-04 | — |
| Input modalities | Text, Image, Video | Text, Image |
| Output modalities | Text | Text |
| Weights | Open weights | Closed / API only |
| Licence | Modified MIT | — |
| Commercial use | Commercial use allowed | Proprietary |
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 Kimi K2.6 or GPT-5 Codex? 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.