cd /news/large-language-models/openai-gpt-oss-120b-mancer-fp8-lt-ch… · home › topics › large-language-models › article
[ARTICLE · art-170211] src=trackllm.net ↗ pub= topic=large-language-models verified=true sentiment=· neutral

openai/gpt-oss-120b @ mancer/fp8: LT change (shift 3.32)

A tracking dashboard recorded a level shift of 3.32 in log-probability drift for the openai/gpt-oss-120b model served at mancer/fp8, while the endpoint's B3IT border-input monitoring was retired on 2026-07-24 for lack of sufficient border inputs. OpenRouter billed $0.074 total for tracking the endpoint since 2026-07-19, including $0.0011 over the last 30 days across 17,148 queries, with B3IT onboarding accounting for $0.06486, LT for $0.01014 and vetting for $0.000591. The drift signals measure how far the endpoint's returned-token log-probabilities and border-input votes have moved from their reference period.

read1 min views2 publishedNov 8, 2026

Tracking status #

LT

tracked— This endpoint is actively tracked. B3IT

retired · no bis— Monitoring was retired: we could not find enough border inputs for this endpoint. (since 2026-07-24)

Drift from baseline #

Both signals measure how far this endpoint's behavior has moved from its reference period. LT tracks the average change in the returned tokens' log-probabilities (nats); B3IT tracks the total-variation distance of the border-input votes (0–1). Both should ideally be as close to 0 as possible. Dashed lines mark published changes, labeled with the level shift across them: for LT the mean drift over the 7 days after minus the 7 days before; for B3IT the total-variation shift its detector measured.

Hover a day on either lane, a change mark or an epoch line to read the data behind it.

Detected changes #

| Date | Method | Level shift |

|---|---|---|

Spend #

what OpenRouter billed for tracking this endpoint. Spend recorded since 2026-07-19 Total to date

$0.074

Last 30 days

$0.0011

Queries

17,148

  • B3IT (onboarding)$0.06486%
  • LT$0.01014%
  • Vetting$0.000591%

Vetting is the one-off probe that checks the endpoint's billed cost against its advertised price.

**B3IT (onboarding)** is the initial border-input search;

**B3IT (monitoring)** and

LT are the recurring daily checks.

── more in #large-language-models 4 stories · sorted by recency
── more on @openai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/openai-gpt-oss-120b-…] indexed:0 read:1min 2026-11-08 · —