gpt-oss 20B

openai/gpt-oss-20b

3.6B active / 21B totalMoEText generationText input

OpenAI · 2 recorded configurations

Measured setup

Hardware

2× RTX 3090 · GPU-resident

Runtime and tool

live llama-server

Quantisation

UD-Q8_K_XL

MTP draft depth

MTP off

Recorded setup

metadata override
No extra overrides

Every option opens a recorded setup. Where the corpus recorded nothing at this page's other settings, the option names the fields it switches with it. Allocation and runtime detail stay in the technical record below.

PP · left axisTG · right axis

View exact measurements

Exact context depth and measured fresh prefill, bulk prefix, and decode rates for unsloth/gpt-oss-20b-GGUF
Context depth (tokens)	Fresh prefill PP (tok/s)	TG (tok/s)
131,072	1,553.07	57.12

Concurrent-stream measurements

Concurrent-stream throughput at the recorded existing-prefix depth for unsloth/gpt-oss-20b-GGUF
Streams	Context depth (tokens)	Aggregate PP (tok/s)	Aggregate TG (tok/s)	TG / stream (tok/s)	Aggregate all (tok/s)
1	131,072	1,553.07	57.12	57.12	157.19
2	131,072	1,480.30	50.13	25.07	140.66

Technical record and provenance

Run details

Source run ID: rtx3090x2-gpt-oss-20b-ud-q8-k-xl-128k-override-2026-09-15-off
Measured date: 2026-09-15
Hardware: 2× NVIDIA GeForce RTX 3090
Runtime: llamaCpp · version: 0.3.0-dev (build 50, commit fc6545d) built with GNU 15.2.0 for Linux x86_64
Weight placement: gpu-resident
KV-cache precision: Q8_0
Flash Attention: Enabled
Batch size: 512
Microbatch size: 512
Prompt tokens: 512
Generated tokens: 256
Speculative decoding: None (standard decoding)
Measured repetitions: 3
Notes: Automated llama-server KV-depth sweep (off). Existing depth is contextTokens; PP=512, TG=256, warmups=1, repetitions=3, Q8_0 K/V. Arithmetic mean of measured repetitions. PP/TG server durations are aligned to client completion timestamps for concurrent phase spans; aggregate-all includes SSH/HTTP wall time. Synthetic repeated-token prompts. Engine was reused as installed; no build was performed. Runtime GGUF metadata override: gpt-oss.context_length=int:132096. The downloaded model file and existing YaRN settings are unchanged; throughput does not establish model quality beyond its native context. The user stopped the unnecessary per-depth concurrency sweep after two completed points. Only completed three-repetition measurements are exported.
Source article: /benchmarks/
Raw log path: research/benchmarks/logs/rtx3090x2/2026-09-15-gpt-oss-20b-ud-q8-k-xl-128k-override

Compare measured entries