Resume experiment gallery
Same source data. Same prompt. Different models. Each entry below is a complete resume page authored end-to-end by the model it names, so you can compare what each one shipped.
Gold Standard
Recommended resume for general viewers.
GPT-OSS 20B
Generated by openai/gpt-oss-20b, self-hosted via http://192.168.0.214:8000.
Gemma4 E2B
Generated by google/gemma-4-E2B-it, self-hosted via http://192.168.0.214:8000.
Gemma4 26B-A4B
Generated by google/gemma-4-26B-A4B-it, self-hosted via http://192.168.0.214:8000.
Qwen3.5 9B
Generated by Qwen/Qwen3.5-9B, self-hosted via http://192.168.0.214:8000.
Gemma4 12B
Generated by google/gemma-4-12B-it, self-hosted via http://192.168.0.214:8000.
LFM2.5 230M
At 230M params, the model didn't write a resume — it echoed fragments of the raw YAML source (including a stray ">-" syntax token) into malformed HTML. Kept as an honest look at the capability floor for this task, not a site bug.
LFM2.5 1.2B Thinking
A real but heavily truncated resume — most experience entries, the summary, community section, and education are missing, and one closing line is fabricated with no basis in the source data. Kept as an honest look at this model's limits at this size, not a site bug.
LFM2.5 8B-A1B
A comprehensive, faithful resume — all roles, projects, and skills accurate, no fabrications. Reproducibly malforms four date tags (confirmed across 3 samples, including with the model's own recommended sampling params) and omits Education & Honors. Kept as an honest look at this model's edges, not a site bug.
Gemma4 E4B
Generated by google/gemma-4-E4B-it, self-hosted via http://192.168.0.214:8000.
Qwen3.5 122B-A10B
Generated by bleysg/Qwen3.5-122B-A10B-int4-fp8-hybrid, self-hosted via http://192.168.0.214:8000.
poolside/Laguna-S-2.1-NVFP4
Generated by poolside/Laguna-S-2.1-NVFP4, self-hosted via http://192.168.0.214:8000.