Real head-to-head · same prompt, one shot
Claude Opus 5.5 vs Grok 4.7
Anthropic's Opus 5.5 — benched on all 50 one-shot builds, every game playtested. vs Grok 4.6's price, a bigger base model, and it builds games that hold together.
Head-to-head verdict: Grok 4.7 wins 12–8.
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 Claude Opus 5.5 and Grok 4.7, side by side, on 20 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.
Claude Opus 5.5 · Benched on all 50 GoldieBench tasks via a Claude subscription: one-shot builds, real renders, pixel-diff playtests on every game, then vision-judged.
Grok 4.7 · Wired into the Agent OS Grok Build tab (4.7 / 4.6 / 4.5 picker, OpenRouter fallback when the CLI is signed out) and benched on twenty skill-infused game builds, each played through its full gameplay arc on a Metal GPU before the vision judge scored the played frame.
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 ↓
Claude Opus 5.5
Grok 4.7
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
Game
— not attempted —
Game
— not attempted —
Game
— not attempted —
Other
— not attempted —
Where Claude Opus 5.5 beat Grok 4.7
The tasks where I gave Claude Opus 5.5 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Crypt
Game
Claude Opus 5.5 8.0
·
Grok 4.7 2.3
(+5.7)
What I saw: Convincing torch-lit crypt: textured stone walls, arch, cobwebs, glowing held torch and a detailed first-person sword, with a clean HUD. No undead visible mid-play and the lighting is a bit evenly warm, so it's strong but not above the field's 9.0.
Voxelcraft
Game
Claude Opus 5.5 5.8
·
Grok 4.7 2.2
(+3.6)
What I saw: Hotbar, hearts, clock and a night-mob system exist, but the mid-play frame is crude: big flat-coloured blocks with no visible texture, the camera jammed against a tree trunk and leaves, and a cramped, hard-to-read view. Also carries the known pre-dusk burning-zombie glow bug.
Nordiccrypt
Game
Claude Opus 5.5 7.6
·
Grok 4.7 5.5
(+2.1)
What I saw: Warm torch-lit brick corridors with bones, barrels, a rune counter, a minimap and a cohesive Nordic HUD make it a proper first-person crawler. The held torch flame is a solid white cone, the axe-sword and hands are crude flat blocks, and no draugr is in view, well short of the fi…
Raycaster
Game
Claude Opus 5.5 8.3
·
Grok 4.7 6.3
(+2.0)
What I saw: Convincing Wolfenstein mood: textured stone corridors, red eagle banners, moss, torches, a cobbled floor, a guard, a treasure chalice and a first-person gun, with a cohesive HUD and minimap. It is a full three.js scene rather than a true raycaster, and the gun and hand are blocky boxes.
Parachute
Game
Claude Opus 5.5 7.6
·
Grok 4.7 6.0
(+1.6)
What I saw: The exit moment works: a detailed plane, an articulated skydiver, altitude/descent HUD, distance-to-clearing compass and a deploy prompt. The mid-play frame is mostly empty grey-blue sky, and no jungle or clearing is visible yet.
Where Grok 4.7 beat Claude Opus 5.5
The tasks where I gave Grok 4.7 a higher 0–10 score on the same prompt — with the actual commentary from my source guides.
Twilightvale
Game
Grok 4.7 8.3
·
Claude Opus 5.5 5.5
(+2.8)
What I saw: Renders a polished twilight open-world with terrain, water pool, scattered trees/monoliths, relics, minimap, and a functional combat loop — HUD shows score climbing, a relic collected, VITAL dropping from a WARDEN fight, all confirming real progression. Strong and shippable, but …
Rpg
Game
Grok 4.7 8.2
·
Claude Opus 5.5 5.5
(+2.7)
What I saw: Strong, atmospheric top-down RPG with a polished themed HUD (vit/dash/relics/inventory/minimap), 3D low-poly world, enemies and interactive seals — HUD evidence shows gold/score/vit all progressing and combat context living. Loses some points because VIT dropped steadily toward d…
Arcade
Game
Grok 4.7 8.6
·
Claude Opus 5.5 6.0
(+2.6)
· 3D breakout polish
What I saw: A gorgeous 3D breakout with rich lighting, gradient sky dome, gridded arena, minimap, and a polished sci-fi HUD; playtest confirms real motion (walkPx 0.23) with hull/lives/score/bricks all changing and no console errors — clearly at or above the field's best.
Neonblaster
Game
Grok 4.7 8.2
·
Claude Opus 5.5 5.8
(+2.4)
What I saw: Renders a polished 3D neon tunnel-runner with slick HUD, minimap, crosshair, gates, ship engine glow and clear state progression (hull draining, wave advancing, score climbing, banners firing) with no errors. Strong and shippable, but it reads more as an on-rails tunnel-runner th…
Skyrim
Game
Grok 4.7 8.4
·
Claude Opus 5.5 6.2
(+2.2)
· snowy shrine explorer
What I saw: Renders a polished, atmospheric first-person Skyrim-lite with soft-shadowed low-poly pines, standing stones, held sword+shield, a working minimap, and a rich RPG HUD (health/stamina/shards/shrines/altitude). Playtest confirms real movement, altitude climbing, and HP dropping from…
Strengths & weaknesses I logged
Claude Opus 5.5
Strengths
- Frontier-class coding + agentic reasoning (Claude 5 family)
- 1M-token context — reads an entire codebase in one call
- All 50 one-shot builds rendered with zero console errors; all 23 games passed the input playtest
Trade-offs
- Premium pricing ($4/$20 per M) — route the everyday 90% to cheaper lanes
- One-shot builds can still ship logic bugs (e.g. a doom kill counter that miscounts)
Grok 4.7
Strengths
- Averaged 7.15/10 on the same twenty skill-infused game briefs where Grok 4.6 averaged 5.95, winning 14 of the 20 head-to-heads
- Best one-shots on this run: neoncity (8.7), racing (8.7), arcade (8.6), every one played on a real GPU before scoring
- Longer reinforcement learning shows: it holds a 40-50KB single-file spec and wires the controls it advertises
- xAI's own coding numbers moved: CursorBench 4.0 46.3% (from 40.4%) and DeepSWE v1.1 71.0% (from 65.2%)
Trade-offs
- 2 of 20 builds died on load or never moved when played — crypt: hud is not a function; voxelcraft: mesh.computeBoundingSphere is not a function file:///Users/j
- Independent Artificial Analysis Intelligence Index v4.3.2 puts it at 46, seven points behind Claude Fable 5.1 and GPT-6 at 53
- Terminal-Bench 4.0 is the weak spot xAI reports itself: 38.0% at launch
Pricing & context — the spec sheet
| Spec | Claude Opus 5.5 | Grok 4.7 |
|---|---|---|
| Vendor | Anthropic | xAI |
| Context window | 1,000,000 tokens | 500,000 tokens |
| Price | $4 / $20 per M | $2 in / $6 out per M tokens |
| Pricing detail | Anthropic's Opus 5.5 — 1M-token context, listed at $4 input / $20 output per million tokens. Benched through a Claude subscription; game tasks use our skill-infused AAA build prompts. | xAI's September 2026 release, priced exactly like Grok 4.6: $2 per million input tokens and $6 per million output on the standard tier, $4 / $12 on the Fast tier (double the output speed). OpenRouter lists the same model at $1.60 / $4.80. |
| Release | 2026-09 | 2026-09 |
| Bench coverage | 50/50 scored · avg 7.57/10 | 20/20 scored · avg 7.15/10 |
The verdict — which should you pick?
Across 20 scored shared tasks, the averages are essentially tied — Claude Opus 5.5 7.21 vs Grok 4.7 7.15. This isn't the comparison where one wins; it's the comparison where you pick based on context, pricing, and what you're actually trying to ship.
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 Claude Opus 5.5 and Grok 4.7 both into the Agent Operating System and dispatch each from the kanban by task type — hardest agentic builds → Claude Opus 5.5, one-shot arcade, driving and shooter builds where the whole game has to arrive in a single file → Grok 4.7. That's the same setup I run for the 4,000+ founders inside the AI Profit Boardroom.
FAQ — Claude Opus 5.5 vs Grok 4.7
Which is better, Claude Opus 5.5 or Grok 4.7?
On Goldie Bench, Claude Opus 5.5 averages 7.21/10 across the shared tasks, with 0 gold, 0 silver, 1 bronze overall. Grok 4.7 averages 7.15/10, with 0 gold, 2 silver, 3 bronze. Grok 4.7 wins the head-to-head 12–8.
How much does Claude Opus 5.5 cost vs Grok 4.7?
Claude Opus 5.5: Anthropic's Opus 5.5 — 1M-token context, listed at $4 input / $20 output per million tokens. Benched through a Claude subscription; game tasks use our skill-infused AAA build prompts. Grok 4.7: xAI's September 2026 release, priced exactly like Grok 4.6: $2 per million input tokens and $6 per million output on the standard tier, $4 / $12 on the Fast tier (double the output speed). OpenRouter lists the same model at $1.60 / $4.80.
What's the context window for Claude Opus 5.5 vs Grok 4.7?
Claude Opus 5.5 has a 1,000,000 tokens context window. Grok 4.7 has a 500,000 tokens context window.
When should I pick Claude Opus 5.5 over Grok 4.7?
Pick Claude Opus 5.5 for: Hardest agentic builds; Whole-repo reasoning; Frontier one-shots. The trade-off is the weaknesses we logged on the bench: Premium pricing ($4/$20 per M) — route the everyday 90% to cheaper lanes; One-shot builds can still ship logic bugs (e.g. a doom kill counter that miscounts).
When should I pick Grok 4.7 over Claude Opus 5.5?
Pick Grok 4.7 for: One-shot arcade, driving and shooter builds where the whole game has to arrive in a single file; High-volume agent work priced at half what the other frontier models charge; Grok Build inside the Agent OS: pick 4.7 in the tab and everything it writes lands in the workspace. The trade-off is the weaknesses we logged on the bench: 2 of 20 builds died on load or never moved when played — crypt: hud is not a function; voxelcraft: mesh.computeBoundingSphere is not a function file:///Users/j; Independent Artificial Analysis Intelligence Index v4.3.2 puts it at 46, seven points behind Claude Fable 5.1 and GPT-6 at 53; Terminal-Bench 4.0 is the weak spot xAI reports itself: 38.0% at launch.
How does Goldie Bench score Claude Opus 5.5 vs Grok 4.7?
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:
Claude Opus 5.5 vs Fusion Grok 4.7 vs Fusion Claude Opus 5.5 vs Claude Opus 5 Grok 4.7 vs Claude Opus 5 Claude Opus 5.5 vs Hermes MoA Grok 4.7 vs Hermes MoA Claude Opus 5.5 vs GPT-5.6 Sol Grok 4.7 vs GPT-5.6 SolFull model pages: Claude Opus 5.5 · Grok 4.7 · 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