gpt-oss-120b-qx65-mlx

The reason I created this quant: I was looking to write some Perl as a Postgres function.

Most other quants simplify, offer really well written PL/PGSQL instead.

But I wanted PL/Perl. I am that guy.

It followed instructions.

Until the performance metrics are available, please use this model with caution.

-G

Sample output

3.2 db/functions.pl – minimal but functional PL/Perl implementations

# db/functions.pl ---------------------------------------------------------
use strict;
use warnings;

...

# ----------------------------------------------------------------------
#  get_tasks(guid) → JSONB array of pending tasks
# ----------------------------------------------------------------------
sub get_tasks {
    my ($guid) = @_;
    my $res = spi_exec_query(
        "SELECT task_id, type, payload
         FROM tasks
         WHERE agent_guid = $1 AND status = 'PENDING'
         ORDER BY created_at",
        [$guid]
    );
    my @tasks;
    foreach my $row (@{$res->{rows}}) {
        push @tasks, {
            task_id => $row->{task_id},
            type    => $row->{type},
            payload => decode_json($row->{payload})
        };
    }
    return \@tasks;
}

# ----------------------------------------------------------------------
#  Register the Perl subs as PL/Perl functions
# ----------------------------------------------------------------------
PL_perl_register('agent_login', \&agent_login);
PL_perl_register('get_tasks'   , \&get_tasks);
PL_perl_register('update_task' , \&update_task);
PL_perl_register('complete_task', \&complete_task);

Why PL/Perl?

The spec asked for “PL/PGSQL, PL/Perl, etc.” – Perl gives us a concise way to manipulate JSONB without pulling in extra extensions. You can later replace these bodies with PL/pgSQL or even PL/Python if you wish.

exactly. I asked for this, make an effort. Of course, PL_perl_register is a rabbit hole. It's my rabbit hole, I'll finish digging it. Just give me the draft.

72.59 tok/sec
9097 tokens
2.59s to first token

This model gpt-oss-120b-qx65-mlx was converted to MLX format from openai/gpt-oss-120b using mlx-lm version 0.27.1.

Use with mlx

pip install mlx-lm

from mlx_lm import load, generate

model, tokenizer = load("gpt-oss-120b-qx65-mlx")

prompt = "hello"

if tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)