bix-12-q2 | 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

What is the median percentage of parsimony informative sites across fungal gene alignments?

Accepted answer: 3.5%
Domain: phylogenetics

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 2.55M total	3/3 replicate 1 replicate 2 replicate 3 749.7K total
Codex GPT-5.6 Sol	3/3 replicate 1 replicate 2 replicate 3 19.06M total	3/3 replicate 1 replicate 2 replicate 3 1.33M total
Codex GPT-5.6 Luna	3/3 replicate 1 replicate 2 replicate 3 44.88M total	3/3 replicate 1 replicate 2 replicate 3 2.33M total
DeepSeek V4 Pro via Codex	3/3 replicate 1 replicate 2 replicate 3 38.20M total	3/3 replicate 1 replicate 2 replicate 3 4.85M total
DeepSeek V4 Pro via Claude Codesuperseded	3/3 replicate 1 replicate 2 replicate 3 4.25M total	1/3 replicate 1 replicate 2 replicate 3 2.09M total

Execution condition

Galaxy-API code with skills

Codex GPT-5.5

3 primary replicates

Codex GPT-5.5

replicate 1

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKITMAFFTmedian summary

Total tokens: 684.2K
Output: 9.9K

Galaxy history Run trace

Codex GPT-5.5

replicate 2

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKITMAFFTBiopythonmedian summary

Total tokens: 936.1K
Output: 10.3K

Galaxy history Run trace

Codex GPT-5.5

replicate 3

passUDT-only

Answer3.53846153846%

Workflow

Galaxy UDTBUSCOPhyKITMAFFTmedian summary

Total tokens: 931.1K
Output: 9.9K

Galaxy history Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKIT

Total tokens: 1.93M
Output: 10.7K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 2

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTMAFFTPhyKIT

Total tokens: 3.00M
Output: 16.4K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 3

pass

Answer3.5385%

Workflow

UnzipFilter From FilePhykit Alignment BasedCollapse DatasetDatamash summaryMAFFT

Total tokens: 14.12M
Output: 25.4K

Galaxy history Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

pass

Answer3.5385%

Workflow

UnzipCollection Element IdentifiersFilter From Filemore stepsSummary statisticsMAFFT

Total tokens: 18.65M
Output: 40.0K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 2

passUDT + standard tools

Answer3.5385%

Workflow

UnzipCollection Element IdentifiersPhykit Alignment BasedGalaxy UDTMAFFTBUSCO

Total tokens: 9.43M
Output: 36.2K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 3

passUDT + standard tools

Answer3.5385%

Workflow

UnzipCollection Element IdentifiersPhykit Alignment Basedmore stepsMAFFTmedian summary

Total tokens: 16.79M
Output: 35.1K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

pass

Answer3.5385%

Workflow

UnzipBUSCOMAFFTPhyKITmean summarymedian summary

Total tokens: 19.43M
Output: 38.4K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 2

pass

Answer3.5385%

Workflow

UnzipPhykit Alignment BasedCollapse DatasetDatamash summary

Total tokens: 7.91M
Output: 41.5K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 3

pass

Answer3.5385%

Workflow

UnzipPhykit Alignment BasedCollection Element Identifiersmore stepsmedian summarymean summary

Total tokens: 10.86M
Output: 34.3K

Galaxy history Run trace

DeepSeek V4 Pro via Claude CodeSuperseded

3 primary replicates

DeepSeek V4 Pro via Claude Code

replicate 1

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKITMAFFTmean summarymedian summary

Total tokens: 1.32M
Output: 19.2K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKITMAFFTmedian summarymean summary

Total tokens: 1.44M
Output: 15.5K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

passUDT-only

Answer3.5385%

Workflow

Galaxy UDTPhyKITMAFFTmean summarymedian summary

Total tokens: 1.50M
Output: 18.1K

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

Answer3.2692307692307696%

Workflow

Local analysisMAFFTmedian summaryBiopython

Total tokens: 269.9K
Output: 4.8K

Run trace

Codex GPT-5.5

replicate 2

pass

Answer3.5384615384615383%

Workflow

Local analysisBUSCOMAFFTmedian summaryBiopython

Total tokens: 290.6K
Output: 4.9K

Run trace

Codex GPT-5.5

replicate 3

pass

Answer3.5384615384615383%

Workflow

Local analysisMAFFTmedian summary

Total tokens: 189.2K
Output: 4.9K

Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

pass

Answer3.5384615384615383%

Workflow

Local analysis

Total tokens: 417.4K
Output: 5.0K

Run trace

Codex GPT-5.6 Sol

replicate 2

pass

Answer3.5385%

Workflow

Local analysis

Total tokens: 640.6K
Output: 6.2K

Run trace

Codex GPT-5.6 Sol

replicate 3

pass

Answer3.54%

Workflow

Local analysis

Total tokens: 273.6K
Output: 4.7K

Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

pass

Answer3.5384615384615383%

Workflow

Local analysis

Total tokens: 543.7K
Output: 11.4K

Run trace

Codex GPT-5.6 Luna

replicate 2

pass

Answer3.5384615384615385%

Workflow

Local analysis

Total tokens: 1.05M
Output: 17.5K

Run trace

Codex GPT-5.6 Luna

replicate 3

pass

Answer3.538461538461538%

Workflow

Local analysis

Total tokens: 735.2K
Output: 17.8K

Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

pass

Answer3.538461538461538%

Workflow

Local analysisBUSCOMAFFTmedian summarymean summary

Total tokens: 1.36M
Output: 14.8K

Run trace

DeepSeek V4 Pro via Codex

replicate 2

pass

Answer3.5385%

Workflow

Local analysisBUSCOMAFFTmedian summarymean summaryPhyKIT

Total tokens: 1.84M
Output: 16.2K

Run trace

DeepSeek V4 Pro via Codex

replicate 3

pass

Answer3.5385%

Workflow

Local analysisMAFFTBiopythonmedian summarymean summaryPhyKIT

Total tokens: 1.65M
Output: 12.3K

Run trace

DeepSeek V4 Pro via Claude CodeSuperseded

3 primary replicates

DeepSeek V4 Pro via Claude Code

replicate 1

fail

Why it missed

Scientific implementation mismatchThis run counted a different alignment population or treated gaps and ambiguous residues as observed amino-acid states. That raised the fungal median to about 5.44% instead of the PhyKIT-compatible 3.54%.

Answer5.437352%

Workflow

Local analysisBUSCOMAFFTmedian summary

Total tokens: 772.8K
Output: 14.8K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

fail

Why it missed

Answer5.44%

Workflow

Local analysisBUSCOMAFFTBiopythonmedian summary

Total tokens: 710.4K
Output: 10.9K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

pass

Answer3.5384615384615383%

Workflow

Local analysisBUSCOMAFFT

Total tokens: 605.2K
Output: 8.6K

Run trace

bix-12-q2 · 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

Answer3.5385%

Total tokens: 1.13M
Input: 1.12M
Cached input: 1.06M
Output: 8.5K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 2

Correct

Answer3.5385%

Total tokens: 2.18M
Input: 2.16M
Cached input: 2.08M
Output: 13.4K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 3

Correct

Answer3.5385%

Total tokens: 1.69M
Input: 1.68M
Cached input: 1.62M
Output: 12.5K

Run trace Final answer Galaxy history

Open-ended code averaged 0.44M tokens; previous website Galaxy averaged 6.35M; the new Galaxy runs averaged 1.67M, or 3.76× 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