Real head-to-head · same prompt, one shot
Kimi K3 vs Fugu Ultra 1.1
Moonshot's 2.8T flagship — 1M context, tuned for long-horizon agent work. vs Sakana's multi-agent orchestrator, v1.1 — routes experts per request.
Head-to-head verdict: Kimi K3 wins 13–8 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 Fugu Ultra 1.1, side by side, on 24 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.
Fugu Ultra 1.1 · Benched on GoldieBench via Sakana's Responses API (fugu-ultra-v1.1, xhigh reasoning). Game tasks use the skill-infused threejs-game-director prompt plus a controls+graphics fix loop with an anti-regression clamp, judged on a real mid-play frame by the same Opus judge as the field.
Side-by-side on 50 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
Fugu Ultra 1.1
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Page
Visual
Where Kimi K3 beat Fugu Ultra 1.1
The tasks where I gave Kimi K3 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Raycaster
Game
Kimi K3 8.6
·
Fugu Ultra 1.1 2.0
(+6.6)
· textured raycaster polish
What I saw: Strong render: multiple distinct textured wall types (brick, blue stone, glowing tech strips, mossy hedge), textured floor/ceiling, minimap, ceiling glow lights and vignette all working with clean UI and controls. Very polished and clearly on-brief; edges out the field with textu…
Neoncity
Game
Kimi K3 8.7
·
Fugu Ultra 1.1 2.3
(+6.4)
· synthwave night drive
What I saw: Gorgeous on-brief cyberpunk drive: glowing multicolor building windows, cyan/magenta neon road edges, a synthwave sunset moon down the vanishing point, and clean HUD with speed/distance—strongly atmospheric and polished. Minor nitpicks are the somewhat sparse traffic and generic …
Pool
Game
Kimi K3 8.1
·
Fugu Ultra 1.1 4.5
(+3.6)
What I saw: Renders a polished 3D pool table with proper rack, numbered balls, cue stick, aim line, pockets, HUD and physics/audio scaffolding — clearly shippable and on-brief. Weak points: the floating lamp shade looks unlit/misplaced and the rack triangle is slightly loose, keeping it just…
Nordiccrypt
Game
Kimi K3 8.6
·
Fugu Ultra 1.1 5.2
(+3.4)
· atmospheric torch-lit crypt
What I saw: Strong: genuine first-person 3D dungeon with convincing torch glow, warm falloff lighting on stone brickwork, scattered rubble props, and a polished serif Nordic UI with rune-collection goal that fully nails the brief. Minor weakness: the corridor geometry is a bit boxy and props…
Aurora
Visual
Kimi K3 8.7
·
Fugu Ultra 1.1 5.5
(+3.2)
· 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…
Where Fugu Ultra 1.1 beat Kimi K3
The tasks where I gave Fugu Ultra 1.1 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Dogfight
Game
Fugu Ultra 1.1 7.2
·
Kimi K3 3.5
(+3.7)
What I saw: Strong visuals — detailed hero jet with shadow, polished HUD, city skyline and terrain create a real air-combat atmosphere; but the screenshot shows no visible enemies or combat (0/10 kills, no dogfight action), so it reads closer to a flight-sim showcase than a proven shooter.
Twilightvale
Game
Fugu Ultra 1.1 8.6
·
Kimi K3 7.3
(+1.3)
· Atmospheric RPG World
What I saw: Gorgeous cohesive twilight scene with detailed hero holding a weapon, visible enemies (creature + humanoid), wooden bridge, lampposts, cottage, river, pickups and a working minimap/HUD with quest and kill tracker. Strong art direction and populated world with combat framing; only…
Doom
Game
Fugu Ultra 1.1 8.4
·
Kimi K3 8.0
(+0.4)
· 3D demon shooter
What I saw: Strong atmospheric 3D maze with a visible horned demon enemy, working shotgun, minimap with tracked enemies, and active combat ('CLAWED' hit feedback, HP dropped to 082) — polished HUD and lighting. Falls just shy of the field's best; enemies are more Three.js models than true ra…
Parachute
Game
Fugu Ultra 1.1 8.6
·
Kimi K3 8.2
(+0.4)
· combat parachute drop
What I saw: Strong deployed-chute skydiver over a jungle canopy with a polished HUD (altitude, dist-to-H, score) plus active drone enemies, flare combat, and 'THREAT DOWN' kill feedback — it delivers the full jump/steer/land loop AND working combat, edging past a bland walking sim.
Crypt
Game
Fugu Ultra 1.1 7.8
·
Kimi K3 7.4
(+0.4)
What I saw: Strong atmospheric torch-lit crypt with a visible skeleton enemy, working HUD (health/torch/kills/depth), coffins and props — clearly a combat crawler not an empty walk sim. Lighting feels a bit washed-out/bright for a 'torch-lit dungeon' and the mood lacks the deep shadowy dread…
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
Fugu Ultra 1.1
Strengths
- Orchestrates 1-3 expert agents per request and synthesises their answers
- Reported SWE-Bench Pro 73.7 — above Opus 4.8 and GPT-5.5 on Sakana's table
- OpenAI- and Anthropic-compatible API — drop-in for Codex and Claude Code
Trade-offs
- Benched as a partial run until the full 50-task batch completes
- Region-locked: unavailable across the EU/EEA/UK/CH — access depends on where you are
Pricing & context — the spec sheet
| Spec | Kimi K3 | Fugu Ultra 1.1 |
|---|---|---|
| Vendor | Moonshot AI | Sakana AI |
| Context window | 1,048,576 tokens — a full codebase in working memory | 1,000,000-token context window |
| Price | $3 / M in | API · orchestration billed |
| Pricing detail | Launched July 16, 2026. 2.8T-param MoE (Moonshot's quickstart corrected the circulating 2.5T estimate). $3/M input on OpenRouter at launch; included at no extra cost in the Kimi coding plan (`k3` on the coding endpoint). | Sakana's Fugu Ultra v1.1 — a multi-agent system served as a model: an orchestrator routes each request across one to three expert agents and synthesises the answer. Reported (v1.0): SWE-Bench Pro 73.7, LiveCodeBench 93.2, GPQA Diamond 95.5. NOTE: Sakana geo-blocks the EU/EEA/UK/Switzerland pending GDPR compliance — benched from an allowed region. |
| Release | 2026-07-16 | 2026-07 |
| Bench coverage | 50/50 scored · avg 7.89/10 | 23/24 scored · avg 6.94/10 |
The verdict — which should you pick?
Across 23 scored shared tasks, Kimi K3 averaged 8.10/10, beating Fugu Ultra 1.1's 6.94/10 by 1.16 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 Fugu Ultra 1.1 both into the Agent Operating System and dispatch each from the kanban by task type — long-horizon agent runs → Kimi K3, hard, high-stakes coding and reasoning where answer quality beats latency → Fugu Ultra 1.1. That's the same setup I run for the 4,000+ founders inside the AI Profit Boardroom.
FAQ — Kimi K3 vs Fugu Ultra 1.1
Which is better, Kimi K3 or Fugu Ultra 1.1?
On Goldie Bench, Kimi K3 averages 8.10/10 across the shared tasks, with 8 gold, 4 silver, 7 bronze overall. Fugu Ultra 1.1 averages 6.94/10, with 0 gold, 1 silver, 2 bronze. Kimi K3 wins the head-to-head 13–8.
How much does Kimi K3 cost vs Fugu Ultra 1.1?
Kimi K3: Launched July 16, 2026. 2.8T-param MoE (Moonshot's quickstart corrected the circulating 2.5T estimate). $3/M input on OpenRouter at launch; included at no extra cost in the Kimi coding plan (`k3` on the coding endpoint). Fugu Ultra 1.1: Sakana's Fugu Ultra v1.1 — a multi-agent system served as a model: an orchestrator routes each request across one to three expert agents and synthesises the answer. Reported (v1.0): SWE-Bench Pro 73.7, LiveCodeBench 93.2, GPQA Diamond 95.5. NOTE: Sakana geo-blocks the EU/EEA/UK/Switzerland pending GDPR compliance — benched from an allowed region.
What's the context window for Kimi K3 vs Fugu Ultra 1.1?
Kimi K3 has a 1,048,576 tokens — a full codebase in working memory context window. Fugu Ultra 1.1 has a 1,000,000-token context window context window.
When should I pick Kimi K3 over Fugu Ultra 1.1?
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 Fugu Ultra 1.1 over Kimi K3?
Pick Fugu Ultra 1.1 for: Hard, high-stakes coding and reasoning where answer quality beats latency; Agentic workflows in Codex / Claude Code via the drop-in provider config; One-shot builds you want a panel of experts on, not a single model. The trade-off is the weaknesses we logged on the bench: Benched as a partial run until the full 50-task batch completes; Region-locked: unavailable across the EU/EEA/UK/CH — access depends on where you are.
How does Goldie Bench score Kimi K3 vs Fugu Ultra 1.1?
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 Fugu Ultra 1.1 vs Fusion Kimi K3 vs Claude Opus 5 Fugu Ultra 1.1 vs Claude Opus 5 Kimi K3 vs Hermes MoA Fugu Ultra 1.1 vs Hermes MoA Kimi K3 vs GPT-5.6 Sol Fugu Ultra 1.1 vs GPT-5.6 SolFull model pages: Kimi K3 · Fugu Ultra 1.1 · 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