
Real head-to-head · same prompt, one shot
Fugu Ultra vs DeepSeek V4 Pro
Sakana's multi-agent answer to Fusion — frontier ensemble without single-vendor risk. vs DeepSeek's flagship tier — benched head-to-head against its own cheap Flash.
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 Fugu Ultra and DeepSeek V4 Pro, side by side, on 42 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.
Fugu Ultra · Dispatched from Agent OS as the panel-ensemble alternative to OpenRouter Fusion. Bench scored by Claude judge against the same 42 prompts as every other model.
DeepSeek V4 Pro · Benched on all 50 GoldieBench tasks via api.deepseek.com with the same pipeline as the Flash 0731 run, then published as a live side-by-side: goldiebench.com/vs-live/deepseek-flash-vs-pro.html loads both builds of every task in twin panes.
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 ↓
Fugu Ultra
DeepSeek V4 Pro
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Page
Page
Sim
Sim
Sim
Strengths & weaknesses I logged
Fugu Ultra
Strengths
- SWE Bench Pro 73.7 · GPQA-D 95.5 · MRCRv2 93.6 — Sakana's published frontier-tier benchmark scores
- Vendor-agnostic ensemble — opt out of specific providers for compliance / export-control
- OpenAI-compatible API at api.sakana.ai — drop-in for existing tooling
Trade-offs
- Panel orchestration adds latency — even a 'pong' burns ~2k orchestration tokens
- Newer than Fusion; less community calibration on long-tail prompts
DeepSeek V4 Pro
Strengths
- Flagship reasoning tier on the same official API and 1M context as Flash
- Ran the identical 50-prompt set as V4 Flash 0731 — a clean same-vendor A/B
- Reasoning-first: thinks before writing every build
Trade-offs
- Unranked — builds are on the bench but not yet scored by the Opus vision judge
- Slower and pricier per build than Flash — the whole question is whether that buys quality
Pricing & context — the spec sheet
| Spec | Fugu Ultra | DeepSeek V4 Pro |
|---|---|---|
| Vendor | Sakana AI | DeepSeek |
| Context window | 272,000 tokens with the standard rate. Calls exceeding 272K context are billed at the higher 'long-context' rates. | 1,000,000-token context window |
| Price | $5 / 1M input · $30 / 1M output (Fugu Ultra) | API · pro tier |
| Pricing detail | Sakana's multi-agent orchestration: a single API call internally dispatches to multiple frontier models and synthesises the answer. Subscription plans run $20-$200/mo (Standard / Pro / Max); PAYG is $5/M input + $30/M output for Fugu Ultra. Direct competitor to OpenRouter Fusion's panel approach. | DeepSeek's flagship tier, benched on `deepseek-v4-pro` via api.deepseek.com — the exact same 50 one-shot prompts, skill-infused game prompt and pipeline as the V4 Flash 0731 run, so the two runs are directly comparable side by side. DeepSeek's own line on the 0731 Flash refresh is that its post-training now beats the older V4-Pro-Preview — this run tests the current Pro against that claim. |
| Release | 2026-06-15 | 2026-07 |
| Bench coverage | 42/42 scored · avg 7.94/10 | 0/50 scored · avg — |
The verdict — which should you pick?
Not enough scored shared tasks yet for a head-to-head average. The live demos for both are on the matrix above — play them and form your own opinion.
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 Fugu Ultra and DeepSeek V4 Pro both into the Agent Operating System and dispatch each from the kanban by task type — teams that want fusion-class quality but need a different vendor risk profile → Fugu Ultra, checking whether deepseek's pro tier is worth the premium over flash 0731 → DeepSeek V4 Pro. That's the same setup I run for the 4,000+ founders inside the AI Profit Boardroom.
FAQ — Fugu Ultra vs DeepSeek V4 Pro
Which is better, Fugu Ultra or DeepSeek V4 Pro?
On Goldie Bench, Fugu Ultra averages no scored verdicts yet across the shared tasks, with 5 gold, 2 silver, 2 bronze overall. DeepSeek V4 Pro averages no scored verdicts yet, with 0 gold, 0 silver, 0 bronze. Not enough scored shared tasks yet to call a winner.
How much does Fugu Ultra cost vs DeepSeek V4 Pro?
Fugu Ultra: Sakana's multi-agent orchestration: a single API call internally dispatches to multiple frontier models and synthesises the answer. Subscription plans run $20-$200/mo (Standard / Pro / Max); PAYG is $5/M input + $30/M output for Fugu Ultra. Direct competitor to OpenRouter Fusion's panel approach. DeepSeek V4 Pro: DeepSeek's flagship tier, benched on `deepseek-v4-pro` via api.deepseek.com — the exact same 50 one-shot prompts, skill-infused game prompt and pipeline as the V4 Flash 0731 run, so the two runs are directly comparable side by side. DeepSeek's own line on the 0731 Flash refresh is that its post-training now beats the older V4-Pro-Preview — this run tests the current Pro against that claim.
What's the context window for Fugu Ultra vs DeepSeek V4 Pro?
Fugu Ultra has a 272,000 tokens with the standard rate. Calls exceeding 272K context are billed at the higher 'long-context' rates. context window. DeepSeek V4 Pro has a 1,000,000-token context window context window.
When should I pick Fugu Ultra over DeepSeek V4 Pro?
Pick Fugu Ultra for: Teams that want Fusion-class quality but need a different vendor risk profile; Operators avoiding export-controlled providers (Sakana emphasises this in their pitch); Deep-research workflows where ensemble verdicts beat single-model answers. The trade-off is the weaknesses we logged on the bench: Panel orchestration adds latency — even a 'pong' burns ~2k orchestration tokens; Newer than Fusion; less community calibration on long-tail prompts.
When should I pick DeepSeek V4 Pro over Fugu Ultra?
Pick DeepSeek V4 Pro for: Checking whether DeepSeek's pro tier is worth the premium over Flash 0731; Hard single-shot builds where extra reasoning depth may pay off. The trade-off is the weaknesses we logged on the bench: Unranked — builds are on the bench but not yet scored by the Opus vision judge; Slower and pricier per build than Flash — the whole question is whether that buys quality.
How does Goldie Bench score Fugu Ultra vs DeepSeek V4 Pro?
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:
Fugu Ultra vs Fusion DeepSeek V4 Pro vs Fusion Fugu Ultra vs Claude Opus 5 DeepSeek V4 Pro vs Claude Opus 5 Fugu Ultra vs Hermes MoA DeepSeek V4 Pro vs Hermes MoA Fugu Ultra vs GPT-5.6 Sol DeepSeek V4 Pro vs GPT-5.6 SolFull model pages: Fugu Ultra · DeepSeek V4 Pro · 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














































