10x Bench Results
See how LLMs perform at coding the Przeprogramowani.pl website in Astro + React + Tailwind + Cloudflare stack.
117 evaluated attempts • Last updated June 28, 2026
Model Family Rankings
GPT-5.4
5 attemptsvia Codex Desktop (High Effort)
Rate: $1.75 / $14 per 1M
Average: 9.1/10.0
GPT-5.4
5 attemptsvia Codex Desktop (High Effort)
Rate: $1.75 / $14 per 1M
Average
9.1/10.0
GPT-5.3-Codex
10 attemptsvia Codex Desktop (High Effort)
Rate: $1.75 / $14 per 1M
Average: 8.5/10.0
GPT-5.3-Codex
10 attemptsvia Codex Desktop (High Effort)
Rate: $1.75 / $14 per 1M
Average
8.5/10.0
GLM-5.2
5 attemptsvia OpenCode
Rate: $1.4 / $4.4 per 1M
Avg run: $0.43 · total $2.17
Average: 8.5/10.0
GLM-5.2
5 attemptsvia OpenCode
Rate: $1.4 / $4.4 per 1M
Avg run: $0.43 · total $2.17
Average
8.5/10.0
DeepSeek V4 Pro
5 attemptsvia OpenCode
Rate: $0.435 / $0.87 per 1M
Avg run: $0.12 · total $0.62
Average: 8.4/10.0
DeepSeek V4 Pro
5 attemptsvia OpenCode
Rate: $0.435 / $0.87 per 1M
Avg run: $0.12 · total $0.62
Average
8.4/10.0
Qwen 3.6
3 attemptsvia Claude Code (High Effort)
Rate: $0.15 / $1 per 1M
Average: 8.3/10.0
Qwen 3.6
3 attemptsvia Claude Code (High Effort)
Rate: $0.15 / $1 per 1M
Average
8.3/10.0
Claude Fable 5
5 attemptsvia Claude Desktop
Rate: $10 / $50 per 1M
Average: 8.3/10.0
Claude Fable 5
5 attemptsvia Claude Desktop
Rate: $10 / $50 per 1M
Average
8.3/10.0
GPT-5.5
5 attemptsvia Codex Desktop (High Effort)
Rate: $5 / $30 per 1M
Average: 7.8/10.0
GPT-5.5
5 attemptsvia Codex Desktop (High Effort)
Rate: $5 / $30 per 1M
Average
7.8/10.0
Gemini 3.5 Flash
5 attemptsvia OpenCode
Rate: $1.5 / $9 per 1M
Average: 7.5/10.0
Gemini 3.5 Flash
5 attemptsvia OpenCode
Rate: $1.5 / $9 per 1M
Average
7.5/10.0
Claude Sonnet 4.6
5 attemptsvia Claude Code (High Effort)
Rate: $3 / $15 per 1M
Average: 7.1/10.0
Claude Sonnet 4.6
5 attemptsvia Claude Code (High Effort)
Rate: $3 / $15 per 1M
Average
7.1/10.0
Claude Opus 4.7
5 attemptsvia Claude Desktop
Rate: $5 / $25 per 1M
Average: 7.0/10.0
Claude Opus 4.7
5 attemptsvia Claude Desktop
Rate: $5 / $25 per 1M
Average
7.0/10.0
Minimax M2.5
5 attemptsvia OpenCode
Rate: $0.3 / $2.4 per 1M
Average: 6.9/10.0
Minimax M2.5
5 attemptsvia OpenCode
Rate: $0.3 / $2.4 per 1M
Average
6.9/10.0
Gemini 3.1 Pro
5 attemptsvia Cursor
Rate: $2 / $12 per 1M
Average: 6.7/10.0
Gemini 3.1 Pro
5 attemptsvia Cursor
Rate: $2 / $12 per 1M
Average
6.7/10.0
Grok Code Fast 1
5 attemptsvia OpenCode
Rate: $0.2 / $1.5 per 1M
Average: 5.9/10.0
Grok Code Fast 1
5 attemptsvia OpenCode
Rate: $0.2 / $1.5 per 1M
Average
5.9/10.0
Qwen 3 Max
3 attemptsvia OpenCode
Rate: $1.2 / $6 per 1M
Average: 4.5/10.0
Qwen 3 Max
3 attemptsvia OpenCode
Rate: $1.2 / $6 per 1M
Average
4.5/10.0
Devstral 2
3 attemptsvia OpenCode
Rate: $0.4 / $2 per 1M
Average: 1.7/10.0
Devstral 2
3 attemptsvia OpenCode
Rate: $0.4 / $2 per 1M
Average
1.7/10.0
Detailed Comparison
Review model averages first, switch to all attempts for scoring notes, or compare two selected models below.
Score Legend
| Criterion | GPT-5.4 5 attempts avg | GLM-5.2 5 attempts avg | GPT-5.3-Codex 10 attempts avg | DeepSeek V4 Pro 5 attempts avg | Qwen 3.6 3 attempts avg | Claude Fable 5 5 attempts avg | GPT-5.5 5 attempts avg | Gemini 3.5 Flash 5 attempts avg | Claude Sonnet 4.6 5 attempts avg | Claude Opus 4.7 5 attempts avg | Minimax M2.5 5 attempts avg | Gemini 3.1 Pro 5 attempts avg | Grok Code Fast 1 5 attempts avg | Qwen 3 Max 3 attempts avg | Devstral 2 3 attempts avg |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Local build | 1 5 attempts | 1 5 attempts | 1 10 attempts | 1 5 attempts | 1 3 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 0.67 3 attempts | 0.33 3 attempts |
Manual testing | 1 5 attempts | 1 5 attempts | 1 10 attempts | 1 5 attempts | 1 3 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 0.60 5 attempts | 0.67 3 attempts | 0.33 3 attempts |
Tech stack | 1 5 attempts | 0.50 5 attempts | 0.90 10 attempts | 0.40 5 attempts | 0.33 3 attempts | 0 5 attempts | 0.40 5 attempts | 0.20 5 attempts | 0.10 5 attempts | 0.10 5 attempts | 0.90 5 attempts | 0.70 5 attempts | 0.30 5 attempts | 0.17 3 attempts | 0 3 attempts |
O nas page | 0.90 5 attempts | 0.50 5 attempts | 0.55 10 attempts | 1 5 attempts | 0.83 3 attempts | 0.90 5 attempts | 0.60 5 attempts | 0.60 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.40 5 attempts | 0.50 5 attempts | 0.33 3 attempts | 0.17 3 attempts |
Podcast page | 0.70 5 attempts | 0.50 5 attempts | 0.70 10 attempts | 0.50 5 attempts | 0.67 3 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.40 5 attempts | 0.50 5 attempts | 0.33 3 attempts | 0.17 3 attempts |
YouTube page | 1 5 attempts | 1 5 attempts | 1 10 attempts | 0.90 5 attempts | 0.50 3 attempts | 0.90 5 attempts | 0.90 5 attempts | 0.70 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.70 5 attempts | 0.40 5 attempts | 0.90 5 attempts | 0.67 3 attempts | 0.17 3 attempts |
Kursy section | 0.70 5 attempts | 1 5 attempts | 0.65 10 attempts | 1 5 attempts | 1 3 attempts | 1 5 attempts | 0.60 5 attempts | 0.60 5 attempts | 0.60 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.33 3 attempts | 0.17 3 attempts |
Consistent UI | 1 5 attempts | 1 5 attempts | 1 10 attempts | 1 5 attempts | 1 3 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 1 5 attempts | 0.80 5 attempts | 0.80 5 attempts | 0.60 5 attempts | 0.67 3 attempts | 0.33 3 attempts |
Responsive design | 0.80 5 attempts | 1 5 attempts | 0.95 10 attempts | 1 5 attempts | 1 3 attempts | 1 5 attempts | 0.90 5 attempts | 1 5 attempts | 1 5 attempts | 0.90 5 attempts | 0.50 5 attempts | 1 5 attempts | 0.40 5 attempts | 0.33 3 attempts | 0 3 attempts |
SEO Tags | 1 5 attempts | 1 5 attempts | 0.75 10 attempts | 0.60 5 attempts | 1 3 attempts | 1 5 attempts | 0.90 5 attempts | 0.90 5 attempts | 0.90 5 attempts | 1 5 attempts | 0.50 5 attempts | 0.50 5 attempts | 0.60 5 attempts | 0.33 3 attempts | 0 3 attempts |
Penalty | 0 5 attempts | 0 5 attempts | N/A0 attempts | 0 5 attempts | 0 3 attempts | 0 5 attempts | 0 5 attempts | 0 5 attempts | 0 5 attempts | 0 5 attempts | N/A0 attempts | 0 5 attempts | N/A0 attempts | N/A0 attempts | N/A0 attempts |
Dots show the individual attempts that make up each model average.
Model Comparison
Compare two model families by average criterion score and attempt spread.
GPT-5.4
9.1/10.0 avg across 5 attempts
Overall delta
A +6.0 pts
GPT-5.3-Codex
8.5/10.0 avg across 10 attempts
| Criterion | GPT-5.4 | Difference | GPT-5.3-Codex |
|---|---|---|---|
| Local build | 1 | Tie | 1 |
| Manual testing | 1 | Tie | 1 |
| Tech stack | 1 | A +0.10 | 0.90 |
| O nas page | 0.90 | A +0.35 | 0.55 |
| Podcast page | 0.70 | Tie | 0.70 |
| YouTube page | 1 | Tie | 1 |
| Kursy section | 0.70 | A +0.05 | 0.65 |
| Consistent UI | 1 | Tie | 1 |
| Responsive design | 0.80 | B +0.15 | 0.95 |
| SEO Tags | 1 | A +0.25 | 0.75 |
| Penalty | 0 | N/A | N/A |