The industry standard AI Gateway for models, mcps and agents https://nitter.cf/t.co/UV2Ppapiyz
try it today ➜
Joined December 2022
- Tweets1.3K
- Following195
- Followers6.2K
- Likes1.1K
Track your Kimi K3.1 Pro spend on @TogetherAI, @FireworksAI_HQ and @baseten using LiteLLM AI Gateway
And here are a few highlighted updates we merged today:
- Added Bedrock GovCloud rates for Kimi, Haiku, Grok, and service tiers (#50128, #50121, #50117).
- Maintained regional & speed multipliers on custom per-call pricing (#50063).
- Standardized batch cached token billing when explicit batch prices are missing (#50062).
- Fixed custom video provider billing to charge once on first completed poll (#49824).
Hello Singapore! 🇸🇬
Some of the @LiteLLM team is in Singapore this week 🇸🇬
If you’re building with LiteLLM here, we’d love to meet and hear what you’re working on. Anyone around?
Lets Meet buff.ly/a1Hmtdy
Did you know you can translate speech live between languages through the LiteLLM Gateway?
Try out gpt-realtime-translate available from @OpenAI and @Azure. Sessions are billed per second of audio sent.
Learn more:
docs.litellm.ai/docs/realtim…
Also merged in the last 24 hours:
• #49783 - Use Claude prompt caching from apps built for OpenAI
• #49708 - Let Claude read a pdf from a web link
• #49789 - Show which model replaces one that is being retired
• #49965 - Stop slow spend-log lookups in the admin dashboard after 30 seconds
Introducing LiteLLM’s Unified Cost Accounting Engine
- Single source of truth for budgets and reports
- Dedicated workers to handle heavy analytics
- Beautiful visualizations for complete cost transparency
Read more: docs.litellm.ai/blog/cost-in…
New model releases, pricing, and benchmark results are all in one place: models.litellm.ai
Introducing the @LiteLLM Model Registry.
Track new model releases, provider deprecations, and benchmark results.
Browse the website or use our RSS feeds to trigger automations.
Check it out → models.litellm.ai
How did we improve LLM translation or cost tracking for users in the last 24 hours?
1 - Codex sessions on Claude now starts with 30.8K tokens, down from 688K (#48972)
2 - PDFs from Claude apps now work on Azure OpenAI (#47166)
3 - PDF token estimates on Azure GPT now match the bill: 141, down from 1664 (#47554)
4 - Fireworks models now show real spend, not $0 (#48900)
Autorouter SDK is here!
Introducing the Autorouter SDK in @LiteLLM
Pick the right model for every prompt across your agents, entirely in Python.
You can:
• Classify prompts with heuristics, an LLM, or custom code
• Map any tiers to any models
• Mix providers, pin sessions, escalate on stalls
No proxy required
20x latency reduction by December 1st, that's where Rust migration is headed.
Read more: docs.litellm.ai/blog/rust-mi…
LiteLLM (YC W23) retweeted
Today, we're launching @LiteLLM Private Registry
With AI agents, public security patches can give attackers a blueprint to exploit vulnerabilities.
For the most severe vulnerabilities, LiteLLM Private Registry gives organizations time to upgrade before we publish public patches
Working on red teaming or rigorous security tests?
Use abliterated models from Refuseless (@refuselessai) and Abliteration (@abliteration_ai) on LiteLLM.
We just added support for Refuseless (@refuselessai) and Abliteration (@abliteration_ai) on @LiteLLM AI Gateway
Read more below ↓
TLDR: What makes GPT-6.1 Terra special?
- Fast computer use at an even lower price
- $1 / $5 per 1M tokens, 50% below GPT-5.6 Terra
- Matches GPT-6 Astra on Agents’ Last Exam at 1/10 the price per token
- Finishes tasks 3x+ faster than Astra and GPT-6.1 Sol
And Day0 support on LiteLLM: models.litellm.ai/models/gpt…
OpenCode Go (@opencode) is now officially supported in LiteLLM.
- Route Go subscription models through your existing LiteLLM setup
- LiteLLM selects the right upstream API, preserves the conversation’s session, and records estimated usage under OpenCode Go.
docs.litellm.ai/blog/opencod…
We have completed our investigation into the attack chain behind the v1.107.0 release. Read it here: docs.litellm.ai/blog/securit…
Hello,
At 6:32 a.m. PDT we became aware of malicious LiteLLM 1.107.0 artifacts published to PyPI, Docker Hub, and GHCR, which were published at approximately 4:40 a.m. PDT. Enterprise deployments pulling only from our private registry are unaffected by this incident.
We are coordinating communication around this incident in this GitHub issue (github.com/BerriAI/litellm/i…).
We have removed the malicious releases and images. If you installed or ran them, please follow the remediation guidance in the above-linked issue and treat credentials accessible to those deployments as exposed.
Sincerely,
Oliver Jensen
Director of Security - LiteLLM
![[SECURITY] litellm 1.107.0 compromised (PyPI removed; Docker latest/main-stable removed): cook.pth...](https://pbs.twimg.com/card_img/2117957580840738816/1hJ9PzZa?format=jpg&name=800x419)
LiteLLM (YC W23) retweeted
@LiteLLM now supports cache-aware routing for self-hosted models on Kubernetes.
It keeps each agent session on the same GPU pod, so vLLM can reuse its prompt cache.
Read the blog: litellm.ai/blog?post=cache-a…
LiteLLM (YC W23) retweeted
Signups are now open for @LiteLLM Handoff, our two-week workshop for developers building cloud agents.
You’ll work with the LiteLLM team to build an agent and deploy an internal beta in your own cloud.
The workshop is remote and limited to 50 developers, October 26 – November 9, 2026
Claude Code + your favorite model?
Just use LiteLLM.
Using our harnesses with other models is supported: just use a proxy like litellm (see our docs page, or ask Claude to set it up for you).
One challenge is there’s a lot of model-specific tool design and prompting that goes into building a good harness. The team and I spend a lot of time dialing these in for each model.
LiteLLM (YC W23) retweeted
Which model names (-m) can you use in Harbor? Now there's one answer: LiteLLM's.
harbortoolkit.com/model
