/v1/completions accepts n and silently returns one choice; /v1/chat/completions refuses n != 1 with 400 #682

Before you start

• I have read the FAQ and my problem is not answered there.
• I have read the Roadmap and this is not already planned there.
• I have searched existing issues and found no duplicate.
• I am on the latest release, or on a freshly rebuilt main when building from source.

What happened

CompletionRequest declares n, and /v1/completions answers n=2 (or any n) with HTTP 200 and a single choice. /v1/chat/completions handles the same field by refusing it: 400 "Only n=1 is supported". A client that asks the completions route for several samples (best-of selection, self-consistency, eval harnesses that pass n) silently gets one and has no way to tell. Reproduced 8/8 on main 377a3bc (n=2 and n=3, greedy and temperature 1.0, streamed and not); the code is unchanged on 74a3877.

Reproduction

curl -s localhost:1941/v1/completions -H 'Content-Type: application/json' -d '{"model": "gpt-oss-20b", "prompt": "Write one word that rhymes with cat:", "max_tokens": 6, "n": 2, "temperature": 1.0}'
# {"id":"cmpl-29c9...","object":"text_completion","created":...,"model":"gpt-oss-20b",
#  "choices":[{"index":0,"text":" Possibly not provided: \"That","finish_reason":"length","logprobs":null}],
#  "usage":{"prompt_tokens":9,"completion_tokens":6,"total_tokens":15}}          <- one choice, usage of one

curl -s localhost:1941/v1/chat/completions -H 'Content-Type: application/json' -d '{"model": "gpt-oss-20b", "max_tokens": 6, "n": 2, "messages": [{"role": "user", "content": "Write one word that rhymes with cat:"}]}'
# {"error":{"message":"Only n=1 is supported","type":"invalid_request_error","param":"n","code":null}}

repro_completions_n_ignored.py
The attached repro_completions_n_ignored.py (stdlib only) checks six shapes:

PASS  [control] completions n=1                  n=1  HTTP 200  choices [0]
PASS  [control] chat n=2 (documented refusal)    n=2  HTTP 400  choices -  body: '{"error":{"message":"Only n=1 is supported",...}}'
FAIL  completions n=2 greedy                     n=2  HTTP 200  choices [0]
FAIL  completions n=2 temperature 1.0            n=2  HTTP 200  choices [0]
FAIL  completions n=3 temperature 1.0            n=3  HTTP 200  choices [0]
FAIL  completions n=2 stream                     n=2  HTTP 200  choices [0]

How did you install FreeToken

Built from source

FreeToken version

ft --version: freetoken version 0.1.3 git rev-parse --short HEAD: 377a3bc (main, 2026-10-07)

OS

Arch Linux

OS details

Ubuntu 20.04.6 LTS, kernel 5.15.0-138-generic, Python 3.12.8.

GPU and driver

NVIDIA GeForce RTX 4090 24 GB (sm_89), driver 580.126.09, CUDA 13.0.

CPU and system RAM

AMD Ryzen Threadripper PRO 5955WX (16 cores / 32 threads), 440 GiB RAM

Checkpoint

openai/gpt-oss-20b (Hugging Face, revision 6cee5e81ee83917806bbde320786a8fb61efebee)

Command

Full log

Anything else

No response