Touchstone
← Models

gpt-oss-20b

evaluatedLight/fast
Vendor: OpenAIParams: 20BQuant: MXFP4Context: 128kLicense: Apache-2.0Type: local
Capability starfish
green = capability (pass-rate) · blue = operational (vs roster) · dashed = roster average
Gate13/13Coding24/24Agentic12/14C#14/15PHP5/5Speed72.1 tok/sDecisive1.4 steps/benchCost-eff$0.0001/run0
96%
Overall
mean domain rate
Solved
73/100
latest per bench
Gauntlet time
2m
~3s/bench
Speed
72.1
tok/s
Steps/bench
1.4
decisiveness
$ / run
$0.0001
OpenRouter-equiv
Overall rank
#6 of 26
Agentic
#13 of 22
Coding
#3 of 22
Speed (tok/s)
#3 of 22
Decisive
#3 of 22
Cheapest
#2 of 9
Strongest
Gate · 100%
Weakest
Agentic · 86%
Where it fits · live, recomputed from current results

#6 of 26 overall; strongest on Gate (100%); weakest on Agentic (86%); #3/22 for speed (72.1 tok/s); #2/9 cheapest ($0.0001/run). (auto-generated from current results)

curator note: The cheapest, fastest equivalent in the roster. Weakest on hard fixes (5/33 solved) but blazing and near-free — good for triage, gate checks, and high-volume cheap passes.

densefastcheapest
Verdict history · snapshotted daily when it changes
2026-07-08#6 of 26 overall; strongest on Gate (100%); weakest on Agentic (86%); #3/22 for speed (72.1 tok/s); #2/9 cheapest ($0.0001/run). (auto-generated from current results)
2026-07-07#11 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (16%); #17/25 on the supportflow gauntlet; #5/26 for speed (55.9 tok/s); #1/9 cheapest ($0.0021/run). (auto-generated from current results)
2026-07-06#12 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (16%); #17/24 on the supportflow gauntlet; #5/26 for speed (55.9 tok/s); #1/9 cheapest ($0.0021/run). (auto-generated from current results)
2026-07-04#12 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (16%); #17/23 on the supportflow gauntlet; #5/26 for speed (55.9 tok/s); #1/9 cheapest ($0.0021/run). (auto-generated from current results)
2026-07-03#11 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (18%); #16/22 on the supportflow gauntlet; #5/26 for speed (56.2 tok/s); #1/9 cheapest ($0.0017/run). (auto-generated from current results)
2026-07-02#8 of 30 overall; strongest on Gate (100%); weakest on SupportFlow (18%); #16/26 on the supportflow gauntlet; #4/29 for speed (56.2 tok/s); #1/9 cheapest ($0.0017/run). (auto-generated from current results)
2026-07-01#10 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (19%); #13/23 on the supportflow gauntlet; #4/27 for speed (54.7 tok/s); #1/9 cheapest ($0.0013/run). (auto-generated from current results)
2026-06-30#10 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (19%); #12/21 on the supportflow gauntlet; #4/27 for speed (54.7 tok/s); #1/9 cheapest ($0.0013/run). (auto-generated from current results)
2026-06-29#9 of 28 overall; strongest on Gate (100%); weakest on SupportFlow (19%); #12/21 on the supportflow gauntlet; #4/27 for speed (54.7 tok/s); #1/9 cheapest ($0.0013/run). (auto-generated from current results)
Domain breakdown
Gate
100%
13/13
Coding
100%
24/24
Agentic
86%
12/14
C#
93%
14/15
PHP
100%
5/5
Benches run
green = passed · red = failed · hover for harness & time
Gate (13)
Coding (24)
Agentic (14)
C# (15)
PHP (5)
SWE (29)
History (monthly pass-rate)
2026-06
85%
137/162
2026-07
98%
114/116
Availability record
When this model appeared in our catalogue census.
2026-06-25 · openrouter2026-06-25 · openrouter