Real head-to-head · same prompt, one shot
Kimi K3 vs Qwen 3.7
Moonshot's 2.5T flagship — 1M context, tuned for long-horizon agent work. vs Multilingual open-weights — strong on Chinese reasoning.
Head-to-head verdict: Kimi K3 wins 29–9.
What I tested — same prompt, two models
I run the same fixed prompt set through every new model the day it drops — same string, one shot, single HTML file out — and I score the result 0–10 on whether it ran, how close it hit the brief, and how good it looked. Below is what came out when I gave the exact same prompts to Kimi K3 and Qwen 3.7, side by side, on 38 shared tasks inside the Agent Operating System.
Both models were given identical prompts inside the Agent Operating System — no help, no iteration, no "best of N" tricks. I run each prompt once, save the HTML file the model produces, and score it 0–10 on whether it ran, how close it hit the brief, and how good it looked. The scoring is mine. The verdicts below are pulled from my source comparison guides at agentos.guide where I publish every score and the reasoning behind it.
Kimi K3 · Wired into the Agent OS as the `kimi-k3` Hermes profile and a K3 speed-toggle in the Kimi Code tab — used for long unattended agent runs where a slow-but-right model beats a fast-but-forgetful one.
Qwen 3.7 · Wired alongside GLM-5.2 in Agent OS for open-weights agent loops where you want vendor diversity.
Side-by-side on 49 shared tasks
Click any cell to play that model's actual one-shot attempt. Medals are derived from my 0–10 scores per task (highest = 🥇, second = 🥈, third = 🥉).
Task ↓
Kimi K3
Qwen 3.7
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Page
Page
Page
Sim
Sim
Sim
Sim
Sim
Sim
Where Kimi K3 beat Qwen 3.7
The tasks where I gave Kimi K3 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Lavalamp
Visual
Kimi K3 8.6
·
Qwen 3.7 5.0
(+3.6)
· metaball wax lamp
What I saw: Strong: a genuinely convincing lava lamp silhouette with glowing base, heated cap, warm wax gradient and a proper metaball shader plus polished typography and theme chips; minor weakness is the blob field looks a touch sparse/stratified in this frame, but the craft and shader det…
Outrun
Game
Kimi K3 8.6
·
Qwen 3.7 5.5
(+3.1)
· Polished synthwave outrun
What I saw: Gorgeous rendered scene — gradient sunset with scanlines, layered mountain ridges, glowing pyramids/palms/signs, and clean pseudo-3D road with dashed lanes and traffic all read perfectly on-brief. Only nit is the player car's flat pink slab looks slightly less refined than the at…
Synthwave
Visual
Kimi K3 9.0
·
Qwen 3.7 6.0
(+3.0)
· textbook synthwave sunset
What I saw: Nails every synthwave trope beautifully — banded sunset, receding neon perspective grid, layered mountains, glowing gradient title, floating wireframe solids, starfield, scanlines/vignette, and generative WebAudio music. Highly polished and cohesive; a clear task winner.
Dragonrealm
Game
Kimi K3 8.6
·
Qwen 3.7 6.0
(+2.6)
· atmospheric frozen realm
What I saw: Strong atmospheric render nails the Skyrim-style frozen night — moon, aurora, snowfall, drawn sword in-hand, glowing braziers, distant dragon silhouette on compass, and a full HUD with health/stamina bars. Polished lighting and moody vignette push it to the top of the field; mino…
Fractal
Sim
Kimi K3 8.6
·
Qwen 3.7 6.0
(+2.6)
· GPU realtime explorer
What I saw: Crisp GPU-shader Mandelbrot renders beautifully with smooth continuous coloring, vivid Aurora palette, and a polished glassy HUD; full feature set (mode toggle, Julia spawn, palettes, zoom/pan, keyboard) makes it a task winner, only mild nit being the intense magenta background d…
Where Qwen 3.7 beat Kimi K3
The tasks where I gave Qwen 3.7 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Orbit
Sim
Qwen 3.7 7.5
·
Kimi K3 3.0
(+4.5)
What I saw: Opus nailed the brief — distinct labelled planet orbits, a real NEO panel, a sim clock. GLM went dramatic with a glowing nebula swirl (gorgeous, but more galaxy than orbit map). Qwen drew a dense, busy orbital swarm — structurally orbit-like but dimmer and harder to read.
Dogfight
Game
Qwen 3.7 7.5
·
Kimi K3 3.5
(+4.0)
What I saw: 18KB · plays clean · webgl, input
Galaxy
Sim
Qwen 3.7 6.0
·
Kimi K3 3.0
(+3.0)
What I saw: 9KB · animation runs but no input response · webgl
Terrain
Visual
Qwen 3.7 7.5
·
Kimi K3 4.5
(+3.0)
What I saw: 4KB · plays clean · webgl, rAF
Solar
Sim
Qwen 3.7 8.0
·
Kimi K3 6.5
(+1.5)
What I saw: 7KB · plays clean · webgl, controls, rAF
Strengths & weaknesses I logged
Kimi K3
Strengths
- Launch-day benchmarks put it around the Fable/Sol tier, with Terminal Bench (agentic terminal-driving) the standout
- 1M-token context verified on this bench's needle test: exact recall from 162k tokens of noise in 18s
- One-shot builds run long but land complete — its first bench game (13.4 min of thinking, 30,880 tokens) playtested with zero JS errors
- Included in the Kimi coding plan — frontier tier without a new bill
Trade-offs
- Slow on hard tasks — early testers report up to ~35 minutes at max reasoning; this bench saw 13+ minute single builds
- Launch-day rate limits on OpenRouter (429s) — the coding-plan endpoint was the reliable route
- Self-reports as K2.7 if you ask it — verify the served model via the API response, not the model's word
Qwen 3.7
Strengths
- Open weights, free for individuals — same model class as GLM-5.2
- Best-of-three on fluid simulation in the Goldie Bench bench
- Multilingual depth — Chinese reasoning especially strong
Trade-offs
- Only 5 tasks scored on the bench so far — small sample size
- Trails GLM-5.2 on cinematic visual builds at similar pricing
Pricing & context — the spec sheet
| Spec | Kimi K3 | Qwen 3.7 |
|---|---|---|
| Vendor | Moonshot AI | Alibaba |
| Context window | 1,048,576 tokens — a full codebase in working memory | 256,000 tokens |
| Price | $3 / M in | Open weights · free for individuals |
| Pricing detail | Launched July 16, 2026. 2.5T-param MoE. $3/M input on OpenRouter at launch; included at no extra cost in the Kimi coding plan (`k3` on the coding endpoint). | Alibaba's open-weights release — downloadable from Hugging Face, runnable locally or via Alibaba Cloud's free tier for individuals. |
| Release | 2026-07-16 | 2026-06 |
| Bench coverage | 40/40 scored · avg 7.79/10 | 47/47 scored · avg 7.00/10 |
The verdict — which should you pick?
Across 38 scored shared tasks, Kimi K3 averaged 7.77/10, beating Qwen 3.7's 7.04/10 by 0.73 points. Pick Kimi K3 when the build has to ship on the first prompt and you can afford the trade-offs in the comparison below.
If you only run one of these inside your stack, the head-to-head average above is the call. If you can run both, my honest play is to wire Kimi K3 and Qwen 3.7 both into the Agent Operating System and dispatch each from the kanban by task type — long-horizon agent runs → Kimi K3, open-weights alternative to glm-5.2 when you want a different model family → Qwen 3.7. That's the same setup I run for the 4,000+ founders inside the AI Profit Boardroom.
FAQ — Kimi K3 vs Qwen 3.7
Which is better, Kimi K3 or Qwen 3.7?
On Goldie Bench, Kimi K3 averages 7.77/10 across the shared tasks, with 9 gold, 8 silver, 7 bronze overall. Qwen 3.7 averages 7.04/10, with 0 gold, 0 silver, 0 bronze. Kimi K3 wins the head-to-head 29–9.
How much does Kimi K3 cost vs Qwen 3.7?
Kimi K3: Launched July 16, 2026. 2.5T-param MoE. $3/M input on OpenRouter at launch; included at no extra cost in the Kimi coding plan (`k3` on the coding endpoint). Qwen 3.7: Alibaba's open-weights release — downloadable from Hugging Face, runnable locally or via Alibaba Cloud's free tier for individuals.
What's the context window for Kimi K3 vs Qwen 3.7?
Kimi K3 has a 1,048,576 tokens — a full codebase in working memory context window. Qwen 3.7 has a 256,000 tokens context window.
When should I pick Kimi K3 over Qwen 3.7?
Pick Kimi K3 for: long-horizon agent runs; whole-repo context work; terminal-driving agents. The trade-off is the weaknesses we logged on the bench: Slow on hard tasks — early testers report up to ~35 minutes at max reasoning; this bench saw 13+ minute single builds; Launch-day rate limits on OpenRouter (429s) — the coding-plan endpoint was the reliable route; Self-reports as K2.7 if you ask it — verify the served model via the API response, not the model's word.
When should I pick Qwen 3.7 over Kimi K3?
Pick Qwen 3.7 for: Open-weights alternative to GLM-5.2 when you want a different model family; Multilingual workloads (Chinese, multi-script content); Fluid and particle simulations. The trade-off is the weaknesses we logged on the bench: Only 5 tasks scored on the bench so far — small sample size; Trails GLM-5.2 on cinematic visual builds at similar pricing.
How does Goldie Bench score Kimi K3 vs Qwen 3.7?
Every demo on this page was built by Julian Goldie inside the Agent Operating System — same fixed prompt for both models, one shot, single HTML file out. Each result gets a 0–10 score on whether it ran, how close it hit the brief, and how good it looked. The highest score on each task gets gold; second gets silver; third gets bronze. See methodology for full provenance.
Related comparisons
Other head-to-heads using the same scoring system:
Kimi K3 vs Fusion Qwen 3.7 vs Fusion Kimi K3 vs Hermes MoA Qwen 3.7 vs Hermes MoA Kimi K3 vs GPT-5.6 Sol Qwen 3.7 vs GPT-5.6 Sol Kimi K3 vs Claude Fable 5 Qwen 3.7 vs Claude Fable 5Full model pages: Kimi K3 · Qwen 3.7 · back to the leaderboard
The same stack Julian uses
Run this stack yourself.
Every demo on this bench was built inside the Agent Operating System — one prompt, one shot, single HTML file out. The Agent OS, the prompts, the templates, the weekly walkthroughs and 4,000+ founders shipping with it every day all live inside the AI Profit Boardroom.
4,000+founders
258documented wins
38countries
$59/momonthly