bix-12-q5 | 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 maximum number of parsimony informative sites in any animal gene alignment?

Accepted answer: 29
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 5.30M total	3/3 replicate 1 replicate 2 replicate 3 679.3K total
Codex GPT-5.6 Sol	3/3 replicate 1 replicate 2 replicate 3 10.69M total	3/3 replicate 1 replicate 2 replicate 3 1.51M total
Codex GPT-5.6 Luna	3/3 replicate 1 replicate 2 replicate 3 28.73M total	3/3 replicate 1 replicate 2 replicate 3 1.46M total
DeepSeek V4 Pro via Codex	3/3 replicate 1 replicate 2 replicate 3 5.59M total	3/3 replicate 1 replicate 2 replicate 3 3.03M total
DeepSeek V4 Pro via Claude Codesuperseded	3/3 replicate 1 replicate 2 replicate 3 6.22M total	1/3 replicate 1 replicate 2 replicate 3 1.23M total

Execution condition

Galaxy-API code with skills

Codex GPT-5.5

3 primary replicates

Codex GPT-5.5

replicate 1

passUDT-only

Answer29

Workflow

Galaxy UDTBUSCOPhyKITMAFFT

Total tokens: 1.62M
Output: 14.7K

Galaxy history Run trace

Codex GPT-5.5

replicate 2

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFTPhyKITBiopython

Total tokens: 1.43M
Output: 14.8K

Galaxy history Run trace

Codex GPT-5.5

replicate 3

passUDT-only

Answer29

Workflow

Galaxy UDTPhyKITBiopythonMAFFT

Total tokens: 2.25M
Output: 10.0K

Galaxy history Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFTPhyKIT

Total tokens: 1.57M
Output: 8.6K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 2

pass

Answer29

Workflow

UnzipCollection Element IdentifiersPhykit Alignment BasedMAFFT

Total tokens: 5.65M
Output: 15.8K

Galaxy history Run trace

Codex GPT-5.6 Sol

replicate 3

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFTPhyKIT

Total tokens: 3.48M
Output: 16.6K

Galaxy history Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

passUDT + standard tools

Answer29

Workflow

UnzipPhykit Alignment BasedGalaxy UDTBUSCOMAFFT

Total tokens: 10.43M
Output: 28.9K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 2

passUDT + standard tools

Answer29

Workflow

UnzipCollection Element IdentifiersGalaxy UDTBUSCOMAFFTPhyKIT

Total tokens: 7.75M
Output: 31.0K

Galaxy history Run trace

Codex GPT-5.6 Luna

replicate 3

passUDT + standard tools

Answer29

Workflow

UnzipFilter From FilePhykit Alignment BasedGalaxy UDTBUSCOMAFFT

Total tokens: 10.55M
Output: 33.4K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFTPhyKIT

Total tokens: 1.17M
Output: 17.7K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 2

passUDT + standard tools

Answer29

Workflow

UnzipPhykit Alignment BasedGalaxy UDTDatamash summaryMAFFT

Total tokens: 3.09M
Output: 21.7K

Galaxy history Run trace

DeepSeek V4 Pro via Codex

replicate 3

passUDT-only

Answer29

Workflow

Galaxy UDTPhyKITMAFFT

Total tokens: 1.33M
Output: 16.0K

Galaxy history Run trace

DeepSeek V4 Pro via Claude CodeSuperseded

3 primary replicates

DeepSeek V4 Pro via Claude Code

replicate 1

passUDT + standard tools

Answer29

Workflow

UnzipGalaxy UDTMAFFTBiopython

Total tokens: 1.58M
Output: 19.5K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFTBiopythonPhyKIT

Total tokens: 1.91M
Output: 20.1K

Galaxy history Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

passUDT-only

Answer29

Workflow

Galaxy UDTMAFFT

Total tokens: 2.73M
Output: 29.9K

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

Answer29

Workflow

Local analysisBUSCOMAFFTBiopython

Total tokens: 288.9K
Output: 6.2K

Run trace

Codex GPT-5.5

replicate 2

pass

Answer29

Workflow

Local analysisBUSCOMAFFT

Total tokens: 204.1K
Output: 4.4K

Run trace

Codex GPT-5.5

replicate 3

pass

Answer29

Workflow

Local analysisBUSCOMAFFT

Total tokens: 186.4K
Output: 4.3K

Run trace

Codex GPT-5.6 Sol

3 primary replicates

Codex GPT-5.6 Sol

replicate 1

pass

Answer29

Workflow

Local analysis

Total tokens: 615.4K
Output: 5.4K

Run trace

Codex GPT-5.6 Sol

replicate 2

pass

Answer29

Workflow

Local analysis

Total tokens: 255.1K
Output: 4.4K

Run trace

Codex GPT-5.6 Sol

replicate 3

pass

Answer29

Workflow

Local analysis

Total tokens: 640.7K
Output: 6.0K

Run trace

Codex GPT-5.6 Luna

3 primary replicates

Codex GPT-5.6 Luna

replicate 1

pass

Answer29

Workflow

Local analysis

Total tokens: 552.8K
Output: 12.6K

Run trace

Codex GPT-5.6 Luna

replicate 2

pass

Answer29

Workflow

Local analysis

Total tokens: 555.5K
Output: 8.3K

Run trace

Codex GPT-5.6 Luna

replicate 3

pass

Answer29

Workflow

Local analysis

Total tokens: 352.3K
Output: 10.2K

Run trace

DeepSeek V4 Pro via Codex

3 primary replicates

DeepSeek V4 Pro via Codex

replicate 1

pass

Answer29

Workflow

Local analysisBUSCOMAFFT

Total tokens: 403.5K
Output: 8.7K

Run trace

DeepSeek V4 Pro via Codex

replicate 2

pass

Answer29

Workflow

Local analysisBUSCOMAFFTPhyKIT

Total tokens: 1.28M
Output: 13.9K

Run trace

DeepSeek V4 Pro via Codex

replicate 3

pass

Answer29

Workflow

Local analysisBUSCOMAFFTmedian summaryIQ-TREE

Total tokens: 1.35M
Output: 12.4K

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's treatment of gaps and ambiguous residues counted 35 informative sites. The PhyKIT-compatible site rules give the accepted maximum of 29.

Answer35

Workflow

Local analysisBUSCOMAFFT

Total tokens: 252.0K
Output: 3.4K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 2

pass

Answer29

Workflow

Local analysisBUSCOMAFFTBiopython

Total tokens: 414.2K
Output: 8.7K

Run trace

DeepSeek V4 Pro via Claude Code

replicate 3

fail

Why it missed

Scientific implementation mismatchThis run's treatment of gaps and ambiguous residues counted 35 informative sites. The PhyKIT-compatible site rules give the accepted maximum of 29.

Answer35

Workflow

Local analysisMAFFTBUSCO

Total tokens: 559.1K
Output: 7.7K

Run trace

bix-12-q5 · 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

Answer29

Total tokens: 2.85M
Input: 2.84M
Cached input: 2.73M
Output: 15.9K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 2

Correct

Answer29

Total tokens: 2.39M
Input: 2.37M
Cached input: 2.28M
Output: 14.3K

Run trace Final answer Galaxy history

Galaxy-API code · October 2026

replicate 3

Correct

Answer29

Total tokens: 2.55M
Input: 2.53M
Cached input: 2.45M
Output: 16.0K

Run trace Final answer Galaxy history

Open-ended code averaged 0.50M tokens; previous website Galaxy averaged 3.56M; the new Galaxy runs averaged 2.60M, or 5.15× 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