July 2026
Foundation + Eval Harness
Different open models can be compared on the same tasks and a reproducible report can be generated.
Planned report: Minimal Open-Model Eval Harness
planned
Open Model Lab
Open Model Lab Timeline
A compact timeline of planned gates, report targets, and status from July 2026 through June 2027.
July 2026 - June 2027
August 2026
SFT Pipeline + Data Quality
The work does not stop at fine-tuning; it shows measurable behavior changes, including regressions.
Planned report: Base vs SFT Behavior Change
planned
September 2026
Preference Optimization / DPO
The report clearly shows where post-training helps and where it creates risk.
Planned report: DPO Behavior Impact
planned
October 2026
Agent Harness / Tool Use
Agent behavior is measured step by step with failure modes, not only success/failure.
Planned report: Post-Training and Agentic Coding Behavior
planned
November 2026
Agent Evals + Long-Horizon Tasks
A reliable evaluation system categorizes why the agent fails.
Planned report: Coding Agent Failure Taxonomy
planned
December 2026
Safety / Red Teaming / Refusal Quality
The model's behavior is measured on both risky and harmless requests.
Planned report: Refusal and Jailbreak Evaluation
planned
January 2027
Reasoning Behavior + Process Evaluation
The system can show flawed processes despite correct answers, or identify breakpoints leading to wrong answers.
Planned report: Outcome vs Process Evaluation
planned
February 2027
Monitorability / Interpretability Start
Behavior changes can be tracked not only from outputs but also from model-internal measurements.
Planned report: Failure Prediction Probes
planned
March 2027
Multimodal / UI Understanding / Computer Use
The model's ability to understand visual interfaces is measured task-by-task.
Planned report: UI Understanding Evaluation
planned
April 2027
Training / Inference Systems Efficiency
The main bottlenecks slowing model research infrastructure can be measured and improved.
Planned report: Open-Model Systems Bottlenecks
planned
May 2027
Data Efficiency + Scaling Ladder
The effect of data quality on behavior and compute efficiency is quantified.
Planned report: Score per GPU-Hour
planned
June 2027
Final Integration + Public Portfolio
An open-source project demonstrates an end-to-end open-model research-engineering loop.
Planned report: Final Technical Report
planned