Qwen3-8B forecasting post-training
According to the model card, the model is post-trained from Qwen3-8B on the OpenForesight forecasting dataset.
Open Source Model Profile · nikhilchandak
OpenForecaster-8B
OpenForecaster-8B is an 8.19B-parameter Qwen3 text-generation fine-tune from nikhilchandak. Its model card documents reinforcement-learning post-training on OpenForesight for calibrated future-event forecasting.
- Publisher
- nikhilchandak
- Task
- text-generation
- Model type
- qwen3
- License
- mit
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
OpenForecaster-8B is published by nikhilchandak as a Qwen3-based text-generation model for open-ended forecasting. The captured configuration identifies Qwen3ForCausalLM and Safetensors metadata reports 8,190,735,360 parameters. According to the model card, it is post-trained from Qwen3-8B with reinforcement learning on the OpenForesight dataset and introduced in a paper on scaling open-ended reasoning.
Recorded capabilities
Calibrated forecasting behavior
According to the model card, it provides calibrated confidence estimates when asked explicitly and reasons about uncertainty and future scenarios.
GRPO reward setup
According to the model card, training used GRPO optimizing a joint reward combining accuracy and Brier score.
Documented forecasting prompt
According to the model card, users prompt for likelihood by event and date, with an example asking for dated predictions with probabilities.
Retrieval-aware use note
According to the model card, retrieved information supplied in context can improve predictions, especially for events after the reported cutoff.
Use cases in the source record
- Open-ended future-event forecasting that asks for calibrated likelihoods with explicit confidence prompting.
- Retrieval-grounded prediction workflows that supply recent developments in context for events beyond the reported cutoff.
Limitations and unknowns
- No independent evaluation results were extracted from this record; benchmark figures in the card are publisher claims.
- Knowledge-cutoff information comes from the publisher model card and has not been independently verified by Ethen.
- Provider state is historical snapshot data, not independently refreshed current availability.
- No context-window value was extracted from this record.
Source and provenance
Source: nikhilchandak/OpenForecaster-8B
Captured: Unknown. Processed: 2026-09-07T19:35:59.355567+00:00.
OpenForecaster-8B OpenForecaster-8B is a specialized language model for open-ended forecasting and predicting future events. This model is post-trained from Qwen3-8B using reinforcement learning on the OpenForesight dataset . It was introduced in the paper Scaling Open-Ended Reasoning to Predict the Future . Performance of OpenForecaster-8B on FutureX Performance on FutureX benchmark in July-August 2025 on non-numeric questions (86 Qs): OpenForecaster-8B has a much higher accuracy than 100B+ models. We limit to models released before April 2025 for a fair, equal knowledge cutoff comparison. Model Description OpenForecaster-8B is tra…
F001F002F003F004F005F006F007F009F010F011F013F014F015F016