bix-55-q1 | Current BixBench results

New token-optimization experiment

Three additional GPT-5.6 Sol Galaxy replicates are available separately. Previous results remain unchanged.

View the October 2026 batch and its run records

Question

How many single-copy orthologs are complete and present in all four proteomes after BUSCO analysis with the eukaryota_odb10 lineage?

Accepted answer: 101
Domain: general

Comparison

Pass counts and token totals across the same three replicates.

Agent	Galaxy-API code	Open-ended code
Codex GPT-5.5	3/3 replicate 1 replicate 2 replicate 3 3.86M total	3/3 replicate 1 replicate 2 replicate 3 5.32M total
Codex GPT-5.6 Sol	3/3 replicate 1 replicate 2 replicate 3 4.25M total	2/3 replicate 1 replicate 2 replicate 3 7.97M total
Codex GPT-5.6 Luna	3/3 replicate 1 replicate 2 replicate 3 43.29M total	2/3 replicate 1 replicate 2 replicate 3 13.25M total
DeepSeek V4 Pro via Codex	3/3 replicate 1 replicate 2 replicate 3 11.13M total	3/3 replicate 1 replicate 2 replicate 3 20.26M total
DeepSeek V4 Pro via Claude Codesuperseded	3/3 replicate 1 replicate 2 replicate 3 6.08M total	2/3 replicate 1 replicate 2 replicate 3 11.41M total

Execution condition

Galaxy-API code with skills

Codex GPT-5.5

3 primary replicates

Codex GPT-5.5

replicate 1

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 1.60M
Output: 13.2K

Galaxy history Run trace

Codex GPT-5.5

replicate 2

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 938.3K
Output: 9.3K

Galaxy history Run trace

Codex GPT-5.5

replicate 3

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 1.32M
Output: 13.6K

Galaxy history Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 1.52M
Output: 11.6K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 2

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 1.51M
Output: 10.2K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 3

pass

Answer101

Workflow

Busco

Total tokens: 1.22M
Output: 6.1K

Galaxy history Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

pass

Answer101

Workflow

BuscoFilter TabularTp Grep ToolSelect columnsJoin datasets

Total tokens: 6.05M
Output: 24.2K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 2

passUDT + standard tools

Answer101

Workflow

Galaxy UDTFilter TabularComp1BUSCO

Total tokens: 30.06M
Output: 49.5K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 3

passUDT + standard tools

Answer101

Workflow

Galaxy UDTFilter TabularTp Grep ToolTp Easyjoin ToolBUSCO

Total tokens: 7.18M
Output: 33.3K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

passUDT-only

Answer101

Workflow

Galaxy UDTBUSCO

Total tokens: 3.23M
Output: 24.0K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 2

pass

Answer101

Workflow

Busco

Total tokens: 1.23M
Output: 11.8K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 3

pass

Answer101

Workflow

BuscoFilter TabularSelect columnsComp1

Total tokens: 6.67M
Output: 35.1K

Galaxy history Run trace

DeepSeek V4 Pro via Claude CodeSuperseded

3 primary replicates

DeepSeek V4 Pro via Claude Code

replicate 1

pass

Answer101

Workflow

Busco

Total tokens: 1.49M
Output: 16.8K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

pass

Answer101

Workflow

Busco

Total tokens: 1.94M
Output: 18.6K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

passUDT + standard tools

Answer101

Workflow

Galaxy UDTBusco

Total tokens: 2.65M
Output: 34.5K

Galaxy history Run trace

Execution condition

Open-ended code with skills

Codex GPT-5.5

3 primary replicates

Codex GPT-5.5

replicate 1

pass

Answer101

Workflow

Local analysisBUSCO

Total tokens: 1.50M
Output: 12.3K

Run trace

Codex GPT-5.5

replicate 2

pass

Answer101

Workflow

Local analysisBUSCO

Total tokens: 2.18M
Output: 13.9K

Run trace

Codex GPT-5.5

replicate 3

pass

Answer101

Workflow

Local analysisBUSCO

Total tokens: 1.64M
Output: 10.9K

Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

fail

Why it missed

BUSCO version changed one classificationBUSCO 5.7.1 classified 129 chicken markers as complete single-copy and gave a four-way intersection of 100. The runs reproducing the accepted 101 used a different BUSCO release and classified one additional shared marker as single-copy.

Answer100

Workflow

Local analysis

Total tokens: 2.12M
Output: 7.1K

Run trace

Codex GPT-5.6 Sol

replicate 2

pass

Answer101

Workflow

Local analysis

Total tokens: 2.67M
Output: 8.6K

Run trace

Codex GPT-5.6 Sol

replicate 3

pass

Answer101

Workflow

Local analysis

Total tokens: 3.18M
Output: 9.2K

Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

pass

Answer101

Workflow

Local analysis

Total tokens: 3.79M
Output: 11.8K

Run trace

Codex GPT-5.6 Luna

replicate 2

pass

Answer101

Workflow

Local analysis

Total tokens: 3.31M
Output: 18.4K

Run trace

Codex GPT-5.6 Luna

replicate 3

fail

Why it missed

Answer100

Workflow

Local analysis

Total tokens: 6.15M
Output: 17.0K

Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

pass

Answer101

Workflow

Local analysisBUSCOBiopython

Total tokens: 5.63M
Output: 25.9K

Run trace

DeepSeek V4 Pro via Codex

replicate 2

pass

Answer101

Workflow

Local analysisBUSCOscikit-learnBiopython

Total tokens: 7.74M
Output: 35.1K

Run trace

DeepSeek V4 Pro via Codex

replicate 3

pass

Answer101

Workflow

Local analysisBUSCOscikit-learnBiopython

Total tokens: 6.89M
Output: 42.4K

Run trace

DeepSeek V4 Pro via Claude CodeSuperseded

3 primary replicates

DeepSeek V4 Pro via Claude Code

replicate 1

pass

Answer101

Workflow

Local analysisBUSCOBiopython

Total tokens: 6.15M
Output: 54.0K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

fail

Why it missed

Scientific implementation substitutionThis run replaced BUSCO with a direct pyhmmer approximation and classified only 64 genes as complete single-copy in all four proteomes. BUSCO with eukaryota_odb10 gives the accepted intersection of 101.

Answer64

Workflow

Local analysisBUSCO

Total tokens: 1.84M
Output: 21.5K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

pass

Answer101

Workflow

Local analysisBUSCOBiopython

Total tokens: 3.42M
Output: 33.1K

Run trace

bix-55-q1 · October 2026 · three additional replicates

GPT-5.6 Sol · token optimization

Galaxy-API code, high reasoning, Fast service. Previous runs are retained above.

Galaxy-API code · October 2026

replicate 1

Correct

Answer101

Total tokens: 891.8K
Input: 880.7K
Cached input: 836.1K
Output: 11.1K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 2

Correct

Answer101

Total tokens: 586.1K
Input: 576.9K
Cached input: 529.5K
Output: 9.2K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 3

Correct

Answer101

Total tokens: 887.6K
Input: 872.2K
Cached input: 827.1K
Output: 15.4K

Run trace Final answer Galaxy history

Open-ended code averaged 2.66M tokens; previous website Galaxy averaged 1.42M; the new Galaxy runs averaged 0.79M, or 0.30× Open-ended code. Each mean covers three replicates of this same task with GPT-5.6 Sol.

Cached input is already included in input tokens. Full nested tool/job evidence is in each HF run's galaxy_execution_records.tar.gz, with an accompanying member index.

Optimization overview and all 150 runs