openai
gpt-oss-20b
20.9B parameters · mixture-of-experts · Base · gpt family
Recommended download
Which version should I download?
Default pick: GGUF · Q4_K_M — Balanced size/quality — good default.
Set your rig to confirm it fits and to see if a higher-quality quant runs fully on your hardware.
| Format | Level | Size | Verdict | Est. speed | Quality note | Swarm | Download |
|---|---|---|---|---|---|---|---|
| GGUF | Q4_K_M | 10.83 GB | Set your rig | — | Balanced size/quality — good default. | 2S / 0L webseed OK | |
| GGUF | F16 | 12.85 GB | Set your rig | — | Native MXFP4 experts in an F16 wrapper — the reference gpt-oss artifact. | 2S / 0L webseed OK |
GGUF · Q4_K_M
10.83 GB
- Est. speed
- —
- Swarm
- 2S / 0L webseed OK
Balanced size/quality — good default.
GGUF · F16
12.85 GB
- Est. speed
- —
- Swarm
- 2S / 0L webseed OK
Native MXFP4 experts in an F16 wrapper — the reference gpt-oss artifact.
Want a different quant? Request it on the board →
Runs fully on
Green = the model's best quant fits fully in GPU/unified memory at 8K context. Tap a card for its full "what runs on it" page.
Context
Advertised 128k · usable ≈ 64–128k (measured — RULER-V2)
How we know
Evidence grade: measured — a published evaluation we can cite, with task and length stated.
- RULER V2 (OpenReview) — observed Jul 18, 2026
- gpt-oss model card (arXiv:2508.10925) — observed Jul 18, 2026
RULER-V2 paper (NeurIPS 2025 Workshop): ~57.2 twelve-task average across 8k-128k (per-length breakdowns in figures only; harder bottom-up benchmark, low absolute scores expected). Model card: dense-layer context extended to 131,072 via YaRN.
Reviewed on Jul 18, 2026.
Run it
Runtime completeness (Q4_K_M torrent): llama.cpp ✅ Ollama – vision sidecar –
- – Ollama — runs the GGUF directly; no Modelfile bundled
- – vision sidecar — not a vision model
Runtime verified: generation-verified · llama.cpp b9957
Context / KV 8,192 tokens · FP16 set above the quant table
Generic commands (no rig set). GPU-offload values assume the model fits on your GPU — set your rig for values tuned to your hardware.
llama-server -m gpt-oss-20b-Q4_K_M.gguf -c 8192 -ngl 999
Use llama-cli in place of llama-server for a one-shot prompt.
FROM ./gpt-oss-20b-Q4_K_M.gguf
PARAMETER num_ctx 8192
PARAMETER num_gpu 999
PARAMETER stop "<|return|>"
PARAMETER stop "<|endoftext|>"
PARAMETER stop "<|call|>"
# Chat template: Ollama uses the template embedded in the GGUF (no TEMPLATE directive needed).
ollama create openai-gpt-oss-20b -f Modelfile
ollama run openai-gpt-oss-20b
GGUF is not a first-class vLLM format.
vLLM GGUF support is experimental and single-file only; prefer safetensors/GPTQ/AWQ for production. If you must, pass the .gguf path to `vllm serve` with --load-format gguf on a recent vLLM.
transformers does not load GGUF weights.
GGUF weights run under llama.cpp, Ollama or LM Studio.
Load gpt-oss-20b-Q4_K_M.gguf, set the context length to 8192 tokens. Set GPU offload to Max (all layers).
MLX runs MLX-format weights only (Apple Silicon). This quant is a GGUF build.
GGUF weights run under llama.cpp, Ollama or LM Studio.
Technical details
Chat template
{#-
In addition to the normal inputs of `messages` and `tools`, this template also accepts the
following kwargs:
- "builtin_tools": A list, can contain "browser" and/or "python".
- "model_identity": A string that optionally describes the model identity.
- "reasoning_effort": A string that describes the reasoning effort, defaults to "medium".
#}
{#- Tool Definition Rendering ============================================== #}
{%- macro render_typescript_type(param_spec, required_params, is_nullable=false) -%}
{%- if param_spec.type == "array" -%}
{%- if param_spec['items'] -%}
{%- if param_spec['items']['type'] == "string" -%}
{{- "string[]" }}
{%- elif param_spec['items']['type'] == "number" -%}
{{- "number[]" }}
{%- elif param_spec['items']['type'] == "integer" -%}
{{- "number[]" }}
{%- elif param_spec['items']['type'] == "boolean" -%}
{{- "boolean[]" }}
{%- else -%}
{%- set inner_type = render_typescript_type(param_spec['items'], required_params) -%}
{%- if inner_type == "object | object" or inner_type|length > 50 -%}
{{- "any[]" }}
{%- else -%}
{{- inner_type + "[]" }}
{%- endif -%}
{%- endif -%}
{%- if param_spec.nullable -%}
{{- " | null" }}
{%- endif -%}
{%- else -%}
{{- "any[]" }}
{%- if param_spec.nullable -%}
{{- " | null" }}
{%- endif -%}
{%- endif -%}
{%- elif param_spec.type is defined and param_spec.type is iterable and param_spec.type is not string and param_spec.type is not mapping and param_spec.type[0] is defined -%}
{#- Handle array of types like ["object", "object"] from Union[dict, list] #}
{%- if param_spec.type | length > 1 -%}
{{- param_spec.type | join(" | ") }}
{%- else -%}
{{- param_spec.type[0] }}
{%- endif -%}
{%- elif param_spec.oneOf -%}
{#- Handle oneOf schemas - check for complex unions and fallback to any #}
{%- set has_object_variants = false -%}
{%- for variant in param_spec.oneOf -%}
{%- if variant.type == "object" -%}
{%- set has_object_variants = true -%}
{%- endif -%}
{%- endfor -%}
{%- if has_object_variants and param_spec.oneOf|length > 1 -%}
{{- "any" }}
{%- else -%}
{%- for variant in param_spec.oneOf -%}
{{- render_typescript_type(variant, required_params) -}}
{%- if variant.description %}
{{- "// " + variant.description }}
{%- endif -%}
{%- if variant.default is defined %}
{{ "// default: " + variant.default|tojson }}
{%- endif -%}
{%- if not loop.last %}
{{- " | " }}
{% endif -%}
{%- endfor -%}
{%- endif -%}
{%- elif param_spec.type == "string" -%}
{%- if param_spec.enum -%}
{{- '"' + param_spec.enum|join('" | "') + '"' -}}
{%- else -%}
{{- "string" }}
{%- if param_spec.nullable %}
{{- " | null" }}
{%- endif -%}
{%- endif -%}
{%- elif param_spec.type == "number" -%}
{{- "number" }}
{%- elif param_spec.type == "integer" -%}
{{- "number" }}
{%- elif param_spec.type == "boolean" -%}
{{- "boolean" }}
{%- elif param_spec.type == "object" -%}
{%- if param_spec.properties -%}
{{- "{\n" }}
{%- for prop_name, prop_spec in param_spec.properties.items() -%}
{{- prop_name -}}
{%- if prop_name not in (param_spec.required or []) -%}
{{- "?" }}
{%- endif -%}
{{- ": " }}
{{ render_typescript_type(prop_spec, param_spec.required or []) }}
{%- if not loop.last -%}
{{-", " }}
{%- endif -%}
{%- endfor -%}
{{- "}" }}
{%- else -%}
{{- "object" }}
{%- endif -%}
{%- else -%}
{{- "any" }}
{%- endif -%}
{%- endmacro -%}
{%- macro render_tool_namespace(namespace_name, tools) -%}
{{- "## " + namespace_name + "\n\n" }}
{{- "namespace " + namespace_name + " {\n\n" }}
{%- for tool in tools %}
{%- set tool = tool.function %}
{{- "// " + tool.description + "\n" }}
{{- "type "+ tool.name + " = " }}
{%- if tool.parameters and tool.parameters.properties %}
{{- "(_: {\n" }}
{%- for param_name, param_spec in tool.parameters.properties.items() %}
{%- if param_spec.description %}
{{- "// " + param_spec.description + "\n" }}
{%- endif %}
{{- param_name }}
{%- if param_name not in (tool.parameters.required or []) -%}
{{- "?" }}
{%- endif -%}
{{- ": " }}
{{- render_typescript_type(param_spec, tool.parameters.required or []) }}
{%- if param_spec.default is defined -%}
{%- if param_spec.enum %}
{{- ", // default: " + param_spec.default }}
{%- elif param_spec.oneOf %}
{{- "// default: " + param_spec.default }}
{%- else %}
{{- ", // default: " + param_spec.default|tojson }}
{%- endif -%}
{%- endif -%}
{%- if not loop.last %}
{{- ",\n" }}
{%- else %}
{{- ",\n" }}
{%- endif -%}
{%- endfor %}
{{- "}) => any;\n\n" }}
{%- else -%}
{{- "() => any;\n\n" }}
{%- endif -%}
{%- endfor %}
{{- "} // namespace " + namespace_name }}
{%- endmacro -%}
{%- macro render_builtin_tools(browser_tool, python_tool) -%}
{%- if browser_tool %}
{{- "## browser\n\n" }}
{{- "// Tool for browsing.\n" }}
{{- "// The `cursor` appears in brackets before each browsing display: `[{cursor}]`.\n" }}
{{- "// Cite information from the tool using the following format:\n" }}
{{- "// `【{cursor}†L{line_start}(-L{line_end})?】`, for example: `【6†L9-L11】` or `【8†L3】`.\n" }}
{{- "// Do not quote more than 10 words directly from the tool output.\n" }}
{{- "// sources=web (default: web)\n" }}
{{- "namespace browser {\n\n" }}
{{- "// Searches for information related to `query` and displays `topn` results.\n" }}
{{- "type search = (_: {\n" }}
{{- "query: string,\n" }}
{{- "topn?: number, // default: 10\n" }}
{{- "source?: string,\n" }}
{{- "}) => any;\n\n" }}
{{- "// Opens the link `id` from the page indicated by `cursor` starting at line number `loc`, showing `num_lines` lines.\n" }}
{{- "// Valid link ids are displayed with the formatting: `【{id}†.*】`.\n" }}
{{- "// If `cursor` is not provided, the most recent page is implied.\n" }}
{{- "// If `id` is a string, it is treated as a fully qualified URL associated with `source`.\n" }}
{{- "// If `loc` is not provided, the viewport will be positioned at the beginning of the document or centered on the most relevant passage, if available.\n" }}
{{- "// Use this function without `id` to scroll to a new location of an opened page.\n" }}
{{- "type open = (_: {\n" }}
{{- "id?: number | string, // default: -1\n" }}
{{- "cursor?: number, // default: -1\n" }}
{{- "loc?: number, // default: -1\n" }}
{{- "num_lines?: number, // default: -1\n" }}
{{- "view_source?: boolean, // default: false\n" }}
{{- "source?: string,\n" }}
{{- "}) => any;\n\n" }}
{{- "// Finds exact matches of `pattern` in the current page, or the page given by `cursor`.\n" }}
{{- "type find = (_: {\n" }}
{{- "pattern: string,\n" }}
{{- "cursor?: number, // default: -1\n" }}
{{- "}) => any;\n\n" }}
{{- "} // namespace browser\n\n" }}
{%- endif -%}
{%- if python_tool %}
{{- "## python\n\n" }}
{{- "Use this tool to execute Python code in your chain of thought. The code will not be shown to the user. This tool should be used for internal reasoning, but not for code that is intended to be visible to the user (e.g. when creating plots, tables, or files).\n\n" }}
{{- "When you send a message containing Python code to python, it will be executed in a stateful Jupyter notebook environment. python will respond with the output of the execution or time out after 120.0 seconds. The drive at '/mnt/data' can be used to save and persist user files. Internet access for this session is UNKNOWN. Depends on the cluster.\n\n" }}
{%- endif -%}
{%- endmacro -%}
{#- System Message Construction ============================================ #}
{%- macro build_system_message() -%}
{%- if model_identity is not defined %}
{%- set model_identity = "You are ChatGPT, a large language model trained by OpenAI." %}
{%- endif %}
{{- model_identity + "\n" }}
{{- "Knowledge cutoff: 2024-06\n" }}
{{- "Current date: " + strftime_now("%Y-%m-%d") + "\n\n" }}
{%- if reasoning_effort is not defined %}
{%- set reasoning_effort = "medium" %}
{%- endif %}
{{- "Reasoning: " + reasoning_effort + "\n\n" }}
{%- if builtin_tools %}
{{- "# Tools\n\n" }}
{%- set available_builtin_tools = namespace(browser=false, python=false) %}
{%- for tool in builtin_tools %}
{%- if tool == "browser" %}
{%- set available_builtin_tools.browser = true %}
{%- elif tool == "python" %}
{%- set available_builtin_tools.python = true %}
{%- endif %}
{%- endfor %}
{{- render_builtin_tools(available_builtin_tools.browser, available_builtin_tools.python) }}
{%- endif -%}
{{- "# Valid channels: analysis, commentary, final. Channel must be included for every message." }}
{%- if tools -%}
{{- "\nCalls to these tools must go to the commentary channel: 'functions'." }}
{%- endif -%}
{%- endmacro -%}
{#- Main Template Logic ================================================= #}
{#- Set defaults #}
{#- Render system message #}
{{- "<|start|>system<|message|>" }}
{{- build_system_message() }}
{{- "<|end|>" }}
{#- Extract developer message #}
{%- if messages[0].role == "developer" or messages[0].role == "system" %}
{%- set developer_message = messages[0].content %}
{%- set loop_messages = messages[1:] %}
{%- else %}
{%- set developer_message = "" %}
{%- set loop_messages = messages %}
{%- endif %}
{#- Render developer message #}
{%- if developer_message or tools %}
{{- "<|start|>developer<|message|>" }}
{%- if developer_message %}
{{- "# Instructions\n\n" }}
{{- developer_message }}
{{- "\n\n" }}
{%- endif %}
{%- if tools -%}
{{- "# Tools\n\n" }}
{{- render_tool_namespace("functions", tools) }}
{%- endif -%}
{{- "<|end|>" }}
{%- endif %}
{#- Render messages #}
{%- set last_tool_call = namespace(name=none) %}
{%- for message in loop_messages -%}
{#- At this point only assistant/user/tool messages should remain #}
{%- if message.role == 'assistant' -%}
{#- Checks to ensure the messages are being passed in the format we expect #}
{%- if "content" in message %}
{%- if "<|channel|>analysis<|message|>" in message.content or "<|channel|>final<|message|>" in message.content %}
{{- raise_exception("You have passed a message containing <|channel|> tags in the content field. Instead of doing this, you should pass analysis messages (the string between '<|message|>' and '<|end|>') in the 'thinking' field, and final messages (the string between '<|message|>' and '<|end|>') in the 'content' field.") }}
{%- endif %}
{%- endif %}
{%- if "thinking" in message %}
{%- if "<|channel|>analysis<|message|>" in message.thinking or "<|channel|>final<|message|>" in message.thinking %}
{{- raise_exception("You have passed a message containing <|channel|> tags in the thinking field. Instead of doing this, you should pass analysis messages (the string between '<|message|>' and '<|end|>') in the 'thinking' field, and final messages (the string between '<|message|>' and '<|end|>') in the 'content' field.") }}
{%- endif %}
{%- endif %}
{%- if "tool_calls" in message %}
{#- We need very careful handling here - we want to drop the tool call analysis message if the model #}
{#- has output a later <|final|> message, but otherwise we want to retain it. This is the only case #}
{#- when we render CoT/analysis messages in inference. #}
{%- set future_final_message = namespace(found=false) %}
{%- for future_message in loop_messages[loop.index:] %}
{%- if future_message.role == 'assistant' and "tool_calls" not in future_message %}
{%- set future_final_message.found = true %}
{%- endif %}
{%- endfor %}
{#- We assume max 1 tool call per message, and so we infer the tool call name #}
{#- in "tool" messages from the most recent assistant tool call name #}
{%- set tool_call = message.tool_calls[0] %}
{%- if tool_call.function %}
{%- set tool_call = tool_call.function %}
{%- endif %}
{%- if message.content and message.thinking %}
{{- raise_exception("Cannot pass both content and thinking in an assistant message with tool calls! Put the analysis message in one or the other, but not both.") }}
{%- elif message.content and not future_final_message.found %}
{{- "<|start|>assistant<|channel|>analysis<|message|>" + message.content + "<|end|>" }}
{%- elif message.thinking and not future_final_message.found %}
{{- "<|start|>assistant<|channel|>analysis<|message|>" + message.thinking + "<|end|>" }}
{%- endif %}
{{- "<|start|>assistant to=" }}
{{- "functions." + tool_call.name + "<|channel|>commentary " }}
{{- (tool_call.content_type if tool_call.content_type is defined else "json") + "<|message|>" }}
{{- tool_call.arguments|tojson }}
{{- "<|call|>" }}
{%- set last_tool_call.name = tool_call.name %}
{%- elif loop.last and not add_generation_prompt %}
{#- Only render the CoT if the final turn is an assistant turn and add_generation_prompt is false #}
{#- This is a situation that should only occur in training, never in inference. #}
{%- if "thinking" in message %}
{{- "<|start|>assistant<|channel|>analysis<|message|>" + message.thinking + "<|end|>" }}
{%- endif %}
{#- <|return|> indicates the end of generation, but <|end|> does not #}
{#- <|return|> should never be an input to the model, but we include it as the final token #}
{#- when training, so the model learns to emit it. #}
{{- "<|start|>assistant<|channel|>final<|message|>" + message.content + "<|return|>" }}
{%- else %}
{#- CoT is dropped during all previous turns, so we never render it for inference #}
{{- "<|start|>assistant<|channel|>final<|message|>" + message.content + "<|end|>" }}
{%- set last_tool_call.name = none %}
{%- endif %}
{%- elif message.role == 'tool' -%}
{%- if last_tool_call.name is none %}
{{- raise_exception("Message has tool role, but there was no previous assistant message with a tool call!") }}
{%- endif %}
{{- "<|start|>functions." + last_tool_call.name }}
{{- " to=assistant<|channel|>commentary<|message|>" + message.content|tojson + "<|end|>" }}
{%- elif message.role == 'user' -%}
{{- "<|start|>user<|message|>" + message.content + "<|end|>" }}
{%- endif -%}
{%- endfor -%}
{#- Generation prompt #}
{%- if add_generation_prompt -%}
<|start|>assistant
{%- endif -%}
Stop strings
<|return|>
<|endoftext|>
<|call|>
Evidence & provenance
Source
- Provenance
- https://huggingface.co/openai/gpt-oss-20b
- Revision pin
-
6cee5e81ee83917806bbde320786a8fb61efebee - Manifest
- Present
License
- Name
- apache-2.0
- Commercial use
- yes
- Access
- Open
File hashes (SHA-256)
-
gpt-oss-20b-q4-k-m/LICENSE
58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd -
gpt-oss-20b-q4-k-m/README.md
d4de7b8cc237ec88bfe5cb19bc412e97daf83ab1045e26cd506df7a7cdde9485 -
gpt-oss-20b-q4-k-m/USAGE_POLICY
d6387ef7985019c45db8d9801e6ac5fd9f98f02b9f1c56f8c5af80c3e8f385d0 -
gpt-oss-20b-q4-k-m/gpt-oss-20b-Q4_K_M.gguf
c27536640e410032865dc68781d80a08b98f8db5e93575919af8ccc0568aeb4f -
gpt-oss-20b-f16/LICENSE
58d1e17ffe5109a7ae296caafcadfdbe6a7d176f0bc4ab01e12a689b0499d8bd -
gpt-oss-20b-f16/README.md
d4de7b8cc237ec88bfe5cb19bc412e97daf83ab1045e26cd506df7a7cdde9485 -
gpt-oss-20b-f16/USAGE_POLICY
d6387ef7985019c45db8d9801e6ac5fd9f98f02b9f1c56f8c5af80c3e8f385d0 -
gpt-oss-20b-f16/gpt-oss-20b-F16.gguf
4e4f9cd88d6456e4f389e7262eca4a8d565211e2b22ece9ca7a8556168ff3c66
Explore further
Appears in
Also consider
Local alternative to
Get an email when a better quant fits your rig.
Set your rig once; we'll alert you when a new or better-fitting build lands.
Performance reports
Real-world throughput reported by the community (and scraped sources).
Community
Reviews, sampler presets and community runtime reports. Be the first to contribute.
Community runtime reports
Unofficial, community-submitted "it loads for me" reports. These are not the official verified-working badge — an admin reviews reports before anything is marked officially verified.
Log in to report whether a quant loads in your runtime.
No community reports yet.