AI Potluck
Model components / Fine-tuned / chat models

TinyLlama-1.1B-Chat-v1.0

TinyLlama

A 1.1B instruction-tuned chat model from the TinyLlama project, built on the Llama architecture and trained on 3T tokens for ~90 days on 16 A100-40G GPUs. It became the canonical sub-2B open chat baseline for edge and local-inference research and still pulls ~2.5M Hugging Face downloads per month two years after release.

1.1B chat model (Llama 2 architecture), DPO-aligned finetune of TinyLlama-1.1B (pretrained on 3T tokens). Released 2023/early-2024; v1.0. License Apache-2.0. Verified live on HF June 2026; foundational small-chat model still heavily pulled.

Openness

5 high confidence
5.0
weights
open(Apache-2.0)
base
TinyLlama-1.1B(open, 3T-token pretrain documented)
post-training-data
open(SFT on UltraChat + DPO on UltraFeedback, both public datasets)
code
open(pretraining project public)
license
Apache-2.0(OSI)

Unusually transparent for a chat model: open base, and named public alignment datasets (UltraChat SFT + UltraFeedback DPO). Corrected from 4 to 5 on 2026-07-29, and the post-training evidence moved out of a free-text `post-training` key into `post-training-data` so it carries a value the rubric can read. The old 4 rested on the base pretraining corpus being documented rather than packaged, but this category scores the post-training data -- what the tuner released -- and zephyr scores 5/open_source on these same two datasets while sitting on Mistral-7B, whose corpus is closed outright. Judging TinyLlama by its base while judging zephyr by its post-training set penalized the more open of the two bases. Strongest openness in this batch alongside Nemotron.

Adoption

4 high confidence
4.0

~2.31M HF downloads last month, a workhorse tiny-chat model for edge/testing/CI and educational use. Solidly in the 1-10M monthly-download band.

Capability

1 high confidence
1.0

Intentionally tiny (1.1B, 2023 Llama-2 architecture). Its value is footprint and openness, not capability; scored 1 against a 2026 frontier comparison set. Not built to compete on reasoning/coding suites.

Unchanged since 2026-07-29 (last edited, not re-checked)