BlogLLM Benchmark Scores as Published, Side by Side
Qwen3.8-Max vs Kimi K3.1 Pro: Benchmark Scores Side by Side
Every test both Qwen3.8-Max by Alibaba and Kimi K3.1 Pro by Moonshot AI have a published number for: 13 tests in 6 areas. Each number is as printed, with the page it comes from and the setup it was run under; nothing is ranked or rescored, and two numbers under different setups are two results, not one comparison.
Models Qwen3.8-Max, Kimi K3.1 ProData version 2026111201, as of 2026-11-12
Model
What each developer prints on its own model page; a dash means it prints none. Runs on is the smallest device that holds the model at 4-bit: the developer's words where it prints them, otherwise our estimate from the size.
| Fact | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| Released | 2026-08-03 | 2026-10-23 |
| Context window | 1M | 1M |
| Max output | 131K | - |
| Input | Text, image, video | Text, image |
| Output | Text | Text |
| Weights | Closed | Open (Kimi K3.1 Pro License) |
| Size | - | 2.8T (104B active) |
| Runs on | - | 8 × B300 (about 1.85 TB at 4-bit, estimate) |
Knowledge Work
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| AutomationBench | 27.3%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page |
Agentic Coding
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| DeepSWE v1.1 | 56.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 2 more pages | 75.5%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Claude Code; Dataset revision: v1.1 |
| Terminal-Bench 4.0 | 17%Mistral AI, Mistral Large 4 (preview), checked 2026-10-15 | |
| ProgramBench Almost@1 | 10.5%Z.ai, GLM-5.3 model card, checked 2026-10-15 | |
| Vibe Code Bench |
Science and Math
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| GPQA Diamond | 92.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 | |
| FrontierMath Tier 4 (v2) | ||
| FrontierMath Tier 1-3 (v2) |
Cybersecurity
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| ExploitBench | 28.8%Z.ai, GLM-5.3 model card, checked 2026-10-15 | 74.7%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: ExploitBench official harness; Attempt budget: 300 turns; pass@5 |
General Knowledge and Reasoning
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| Humanity's Last Exam (no tools) | 43.6%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14 | |
| Humanity's Last Exam (with tools) | 56.2%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more pageTools: enabled; exact tools in source | |
| SimpleQA Verified |
Tool Use and Workflows
| Test | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| Agents' Last Exam (pass rate) | 27%Alibaba, Qwen3.8-Max comparison, checked 2026-10-14; printed on 1 more page | 43.2%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Kimi Code; Dataset revision: v1.1 |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked; a dash means the developer lists that price in another unit or not at all. Cost per task is measured by Vals AI.
| Price | Qwen3.8-Max | Kimi K3.1 Pro |
|---|---|---|
| Input | $1.65Alibaba, Alibaba Model Studio pricing, checked 2026-10-15 | $2Moonshot AI, Kimi API pricing (international platform), checked 2026-10-24 |
| Cached input | - | $0.2Moonshot AI, Kimi API pricing (international platform), checked 2026-10-24 |
| Cache write | - | |
| Cache write, 1 hour | - | $4Moonshot AI, Kimi API pricing (international platform), checked 2026-10-24 |
| Output | $4.951Alibaba, Alibaba Model Studio pricing, checked 2026-10-15 | $10Moonshot AI, Kimi API pricing (international platform), checked 2026-10-24 |
| Cost per taskVals AI, USD, lower is better |
Qwen3.8-Max has 33 more and Kimi K3.1 Pro has 42 more: tests only one of them has a number for, listed on its model page.
More side by side
Every number here is read from the same catalog file as the benchmark table, data version 2026111201. A pair has a page when both models are their developers' flagships and both have a number for at least 10 of the same tests. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.