Real head-to-head · same prompt, one shot
Kimi K3 vs Claude Sonnet 5
Moonshot's 2.5T flagship — 1M context, tuned for long-horizon agent work. vs The agentic SWE frontier — 82% SWE-bench Verified, Dev Team mode.
Head-to-head verdict: Kimi K3 wins 28–8 with 2 ties.
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 Claude Sonnet 5, 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.
Claude Sonnet 5 · Reach for it in Agent OS when the job is iterative, tool-using software engineering. For one-shot visual builds, GLM 5.2 (free) beat it 4-1 here.
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
Claude Sonnet 5
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 Claude Sonnet 5
The tasks where I gave Kimi K3 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Aurora
Visual
Kimi K3 8.7
·
Claude Sonnet 5 2.5
(+6.2)
· volumetric 3D aurora
What I saw: Gorgeous flowing volumetric aurora ribbons with convincing fbm noise, layered mountains, spruce silhouettes, moon, stars and a shooting star make a genuinely atmospheric scene; the elegant typography, palette switcher and vignette give it a shippable polish that edges past the fi…
Wormhole
Sim
Kimi K3 8.6
·
Claude Sonnet 5 3.0
(+5.6)
· glowing tunnel depth
What I saw: Gorgeous, on-brief render — concentric glowing gate rings receding into a bright vanishing point, colorful volumetric plasma glow and floating energy particles sell real tunnel depth, with a polished HUD/title. Only nit is the hard diagonal seam in the upper-right that slightly b…
Blackhole
Sim
Kimi K3 9.0
·
Claude Sonnet 5 5.0
(+4.0)
· stunning geodesic lensing
What I saw: Gorgeous real geodesic raytracing with a properly lensed accretion disk wrapping over the top, convincing Doppler asymmetry, photon-ring glow, and a rich nebula starfield backdrop; polished typography and clean UI push it to the top of the field. Minor stat overlap at bottom-left…
Solar
Sim
Kimi K3 6.5
·
Claude Sonnet 5 2.5
(+4.0)
What I saw: Strong UI chrome (glass panels, chips, info card, speed clock) and rich procedural textures, but the rendered view is visually muddy — an oversized washed-out sun bleeds a brownish haze across the whole scene, orbit rings barely read, and the focused 'Mercury' shows a giant gray …
Gtadrive
Game
Kimi K3 8.4
·
Claude Sonnet 5 4.5
(+3.9)
· neon night city
What I saw: Strong atmospheric render — cohesive dusk/neon skyline, lit facade textures, on-foot character with steal prompt, circular minimap with compass, and clean HUD all deliver the GTA sandbox brief. Loses a touch for a slightly empty foreground and unverified cop-chase/wanted mechanic…
Where Claude Sonnet 5 beat Kimi K3
The tasks where I gave Claude Sonnet 5 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Galaxy
Sim
Claude Sonnet 5 8.6
·
Kimi K3 3.0
(+5.6)
· gorgeous spiral swirl
What I saw: Beautiful multi-arm spiral with convincing color gradient (warm core to violet edges), bright glowing bulge, and background starfield—clearly on-brief and polished. Full swirl/orbit/zoom interactivity with a mouse-influence vortex on the particles makes this a task winner.
Terrain
Visual
Claude Sonnet 5 8.0
·
Kimi K3 4.5
(+3.5)
What I saw: Renders a clean, atmospheric procedural landscape with convincing height-based coloring (sand/grass/rock/snow), scattered trees, and fog depth; solid and shippable but the terrain reads somewhat soft/generic and lacks standout visual punch or striking peaks to top the field.
Dogfight
Game
Claude Sonnet 5 6.5
·
Kimi K3 3.5
(+3.0)
What I saw: Renders cleanly with a polished HUD, crosshair, and 3D plane in a proper 3D scene, but the screenshot shows the player facing straight down at empty ground with no enemies visible, and the plane model looks flat/basic — functional but generic compared to the field's best.
Reactiondiff
Sim
Claude Sonnet 5 8.4
·
Kimi K3 6.5
(+1.9)
· GPU Turing patterns
What I saw: Strong GPU Gray-Scott sim rendering clean cell/worm Turing structures with a polished glassy control panel, presets, and interactive seeding; slight weakness is somewhat uniform blob patterns rather than more dramatic branching coral, keeping it just under the top.
Aipbpromo
Page
Claude Sonnet 5 7.5
·
Kimi K3 6.5
(+1.0)
What I saw: 22KB · plays clean · plain
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
Claude Sonnet 5
Strengths
- 82.1% SWE-bench Verified — first model past 80% on real GitHub-issue repair
- Dev Team multi-agent mode + 1M context for repo-level agentic work
- Precision on hard logic — won the raycaster the open-weight field kept botching
Trade-offs
- One-shot creative-visual builds trail GLM 5.2 here (lost 4 of 5) — no iteration to catch its own bugs
- A temporal-dead-zone bug blanked its N-body orbit sim on the first shot
Pricing & context — the spec sheet
| Spec | Kimi K3 | Claude Sonnet 5 |
|---|---|---|
| Vendor | Moonshot AI | Anthropic |
| Context window | 1,048,576 tokens — a full codebase in working memory | 1,000,000 tokens |
| Price | $3 / M in | $3 / $15 per M ($2/$10 intro) |
| 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). | $3.00 input / $15.00 output per million tokens; introductory $2.00/$10.00 through 2026-08-31. |
| Release | 2026-07-16 | 2026-06-30 |
| Bench coverage | 40/40 scored · avg 7.79/10 | 47/47 scored · avg 7.01/10 |
The verdict — which should you pick?
Across 38 scored shared tasks, Kimi K3 averaged 7.77/10, beating Claude Sonnet 5's 7.00/10 by 0.78 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 Claude Sonnet 5 both into the Agent Operating System and dispatch each from the kanban by task type — long-horizon agent runs → Kimi K3, agentic software engineering — write / run / test / fix loops on real repos → Claude Sonnet 5. That's the same setup I run for the 4,000+ founders inside the AI Profit Boardroom.
FAQ — Kimi K3 vs Claude Sonnet 5
Which is better, Kimi K3 or Claude Sonnet 5?
On Goldie Bench, Kimi K3 averages 7.77/10 across the shared tasks, with 9 gold, 8 silver, 7 bronze overall. Claude Sonnet 5 averages 7.00/10, with 1 gold, 1 silver, 5 bronze. Kimi K3 wins the head-to-head 28–8.
How much does Kimi K3 cost vs Claude Sonnet 5?
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). Claude Sonnet 5: $3.00 input / $15.00 output per million tokens; introductory $2.00/$10.00 through 2026-08-31.
What's the context window for Kimi K3 vs Claude Sonnet 5?
Kimi K3 has a 1,048,576 tokens — a full codebase in working memory context window. Claude Sonnet 5 has a 1,000,000 tokens context window.
When should I pick Kimi K3 over Claude Sonnet 5?
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 Claude Sonnet 5 over Kimi K3?
Pick Claude Sonnet 5 for: Agentic software engineering — write / run / test / fix loops on real repos; Repo-level reasoning across a 1M-token context (Dev Team multi-agent mode); Precise logic — raycasters, physics — where one-shot open models slip. The trade-off is the weaknesses we logged on the bench: One-shot creative-visual builds trail GLM 5.2 here (lost 4 of 5) — no iteration to catch its own bugs; A temporal-dead-zone bug blanked its N-body orbit sim on the first shot.
How does Goldie Bench score Kimi K3 vs Claude Sonnet 5?
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 Claude Sonnet 5 vs Fusion Kimi K3 vs Hermes MoA Claude Sonnet 5 vs Hermes MoA Kimi K3 vs GPT-5.6 Sol Claude Sonnet 5 vs GPT-5.6 Sol Kimi K3 vs Claude Fable 5 Claude Sonnet 5 vs Claude Fable 5Full model pages: Kimi K3 · Claude Sonnet 5 · 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