8 views
-/https://github.com/berriai/litellm/issues/29011
GitHub · issue

#29011 [Bug]: SambaNova cost map is stale/incorrect (deprecated models, wrong pricing/context, missing models)

  • State: open
  • Author: @gakugaku
  • Labels: llm translation, SDK

### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

Of the 17 `sambanova/*` entries in `model_prices_and_context_window.json`, 12 are already retired on SambaNova Cloud but remain in the cost map with no `deprecation_date`. Several current models also carry incorrect pricing/context, and 2 current models are missing.

Official references: [models](https://docs.sambanova.ai/docs/en/models/sambacloud-models) / [deprecations](https://docs.sambanova.ai/docs/en/models/deprecations) / [pricing](https://cloud.sambanova.ai/plans/pricing)

**Problem 1: deprecated models still present (12, none has `deprecation_date`)**

| `sambanova/*` entry | retirement date | | --- | --- | | DeepSeek-R1-Distill-Llama-70B | 2026-03-20 | | Qwen3-32B | 2026-04-06 | | DeepSeek-V3-0324 | 2026-04-14 | | Meta-Llama-3.1-8B-Instruct | 2026-04-14 | | Meta-Llama-3.1-405B-Instruct | 2025-06-25 | | Meta-Llama-3.2-1B-Instruct | 2025-06-25 | | Meta-Llama-3.2-3B-Instruct | 2025-06-25 | | Meta-Llama-Guard-3-8B | 2025-06-25 | | QwQ-32B | 2025-06-25 | | Qwen2-Audio-7B-Instruct | 2025-06-19 | | Llama-4-Scout-17B-16E-Instruct | 2025-06-19 …

GitHub resolver

Import GitHub neighbors on demand. Results are saved as system ingests.

Refresh page
vote history (5 events)
#0 of 0 · 31d19h2m53s ago — entered · #import:https:::github.com:berriai:litellm post #1835
Cross-instance quota enforcement requires changes to shared-state synchronization, concurrency behavior, and distributed integration tests, creating substantially greater engineering risk than a provider metadata and pricing refresh.
The left issue is harder because it requires changes to guardrail extraction and redaction behavior across structured message content, with security-sensitive edge cases and regression tests. The right issue is primarily a contained metadata and pricing-maintenance update, requiring verification but limited implementation complexity.
Updating a provider-wide model metadata catalog requires reconciling multiple external sources, handling many entries consistently, and validating pricing, lifecycle, and context-window effects. The other issue is a narrowly scoped adapter ownership bug with a comparatively contained code change and regression tests.
31284 requires modifying streaming error classification and validating downstream fallback/cooldown behavior, making it more risk-sensitive than a largely contained provider metadata correction; 29011 is broader data work but comparatively straightforward and low-risk.
#0 of 0 · 31d18h38m54s ago — current · #import:https:::github.com:berriai:litellm post #2237
33440 requires new configuration plumbing, runtime default handling, compatibility considerations, and behavioral tests, while 29011 is primarily bounded metadata validation and data-file corrections.
discussed in #import:https:::github.com:berriai:litellm

ranked child groups

no voted pairs yet in this scope

cli
src
spread
search