πͺ’ Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
chore(pricing): correct Claude Sonnet 5.5 cache-read price (#18667)
langfuse-bot32ace98b52e0
diff --git a/.agents/skills/add-model-price/references/model-audit-memory.md b/.agents/skills/add-model-price/references/model-audit-memory.mdindex b783fc3c12..545960d8fe 100644--- a/.agents/skills/add-model-price/references/model-audit-memory.md+++ b/.agents/skills/add-model-price/references/model-audit-memory.md@@ -12,11 +12,58 @@ audit date.## Latest useful snapshot-**Audit date:** 2026-09-30+**Audit date:** 2026-10-13All prices listed as `$X / MTok` (per million tokens). Per-token JSON values: divide by 1,000,000.-The 2026-09-30 run re-fetched the full Anthropic pricing page (model table,+The 2026-10-13 run re-fetched the full Anthropic pricing page (model table, cache-hits+footnote, Fast mode, Batch, and tool-use tables) and the Anthropic model-deprecations+page, targeted rows of the OpenAI aggregate Standard pricing table (the full gpt-6/gpt-5.6+family plus every gpt-5.x/gpt-4.x/o-series/legacy model and `gpt-5-chat-latest`'s own+dedicated page), the OpenAI deprecations page and model catalog, the Gemini 3.x and 2.5+pricing pages plus the Gemini models catalog and deprecations page, the AWS Bedrock+pricing page, and the TypeSafe Jev models page. This run found and corrected one stale+price: **Claude Sonnet 5.5's cache-read price dropped from $0.20/MTok (the standard 0.1x+multiplier) to $0.10/MTok (0.05x)**. The Anthropic pricing page's cache-hits footnote now+reads "Cache hits and refreshes on Claude Opus 5.5 and Claude Sonnet 5.5 are priced at+0.05x the base input price" (previously only Opus 5.5 had a non-standard multiplier; Sonnet+5.5 was previously 0.1x like every other non-Fable-5.1/Mythos-5.1/Opus-5.5 model). The AWS+Bedrock pricing page independently corroborates this with a footnote: "Starting October 7,+2026, cache read pricing for Claude Sonnet 5.5 is reduced by 50% from the prior rate" β+Bedrock could not be queried for the exact numeric row (see the standing Bedrock+table-extraction limitation below), but the 50% reduction off $0.20 matches the new $0.10+Anthropic-page rate exactly, so the correction was applied with high confidence from the+direct Anthropic source alone. Updated `claude-sonnet-5-5`'s `cache_read_input_tokens` /+`input_cached_tokens` / `input_cache_read` from `2e-7` to `0.1e-6` in the pricing file; every+other field (input, output, cache-write 5m/1h) is unchanged. Every other price already in+the file β including every `gpt-6-astra`/`gpt-6-sol`/`gpt-6.1-sol`/`gpt-6-luna`,+`gpt-5.6-sol`/`terra`/`luna`, the full `gpt-5.x`/`gpt-4.x`/o-series/legacy OpenAI catalog,+`gpt-5-chat-latest`, `claude-opus-5-5`, `claude-haiku-5-5`, and `gemini-3.6/3.8-flash`/+`3.5-flash-lite` tier β matched verbatim; no other updates were needed. The OpenAI+deprecations page confirms no new deprecation was announced between October 9 and October+13, 2026, and the "Legacy GPT model snapshots" batch (`gpt-3.5-turbo-0125`, `gpt-4-0613`,+`gpt-4-1106-preview`, `gpt-4-turbo`/`gpt-4-turbo-2024-04-09`, `gpt-4.1-nano`,+`gpt-4o-2024-05-13`, `o1`/`o1-2024-12-17`, `o1-pro`/`o1-pro-2025-03-19`,+`o3-mini`/`o3-mini-2025-01-31`, `o4-mini`/`o4-mini-2025-04-16`) still shuts down October 23,+2026 β still in the future as of this run, so no `openAIModels` removal yet; re-check on or+after that date per the 2026-10-09/2026-10-12 entries below. The Anthropic deprecations page+confirms `claude-sonnet-4-5-20250929` is still only `Deprecated` (retirement November 30,+2026, not yet passed) and `claude-haiku-4-5-20251001`'s tentative retirement+("Not sooner than October 15, 2026") has not been formally announced with a firm date β+both left unchanged in `anthropicModels`. The Gemini deprecations page confirms+`gemini-2.5-pro`/`gemini-2.5-flash`/`gemini-2.5-flash-lite` still have "No shutdown date+announced" (access-restriction only, not a deprecation), and `gemini-2.0-flash-thinking-exp-01-21`/+`gemini-1.5-pro`/`gemini-1.5-flash`/`gemini-1.5-flash-8b` are still absent from both the+Gemini models catalog and the deprecations page (neither confirmed available nor shut+down) β standing ambiguous gap, left unchanged in `googleAIStudioModels`. One new fetch+limitation found this run: targeted single-model WebFetch queries against+`ai.google.dev/gemini-api/docs/pricing` for `gemini-3.7-flash` and `gemini-3.5-flash`+(the non-Lite, non-3.6/3.8 siblings) each returned "model not found on this page" even+though the models catalog page lists both as active (not shut down) β see the new+"Gemini 3.x pricing-page single-model fetch gap" entry in+`provider-sources-and-price-keys.md`. Treated as a fetch/extraction artifact, not evidence+of removal; both prices left unchanged and not marked independently re-confirmed this run.+The 2026-09-30 run before it re-fetched the full Anthropic pricing page (model table,cache-hits footnote, Fast mode and Batch tables), the Anthropic models-overviewcomparison table, the OpenAI aggregate Standard/Long-Context/Fast-mode/Flexpricing tables, the full OpenAI model catalog, both Gemini pricing pages (3.x@@ -106,7 +153,7 @@ to have siblings `gpt-5.5-cyber`/`gpt-5.4-cyber`, see provider-sources-and-price| Anthropic | claude-opus-4-1-20250805 | Input $15/MTok, Output $75/MTok, 5m $18.75/MTok, 1h $30/MTok, read $1.50/MTok | Yes | Deprecated β no tiering | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing https://platform.claude.com/docs/en/about-claude/model-deprecations | Still "retired, except on Bedrock and Google Cloud"; deprecations page now shows a firm retirement date of August 5, 2026 (previously open-ended). || Anthropic | claude-opus-4-20250514 | Input $15/MTok, Output $75/MTok, 5m $18.75/MTok, 1h $30/MTok, read $1.50/MTok | Yes | Retired except Google Cloud β no tiering | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing https://platform.claude.com/docs/en/about-claude/model-deprecations | Re-confirmed present; deprecations page shows firm retirement date June 15, 2026. || Anthropic | claude-sonnet-5 | Input $2/MTok, Output $10/MTok; 5m $2.50/MTok, 1h $4/MTok, read $0.20/MTok | Yes | Flat 1M context; permanent pricing (no Sep 1 2026 increase) | Yes | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed unchanged; the $2/$10 permanent-pricing note is still on the page verbatim. Now "legacy (still available)" behind Sonnet 5.5. |-| Anthropic | claude-sonnet-5-5 | Input $2/MTok, Output $10/MTok; 5m $2.50/MTok, 1h $4/MTok, read $0.20/MTok (standard 0.1x multiplier) | Yes | Flat 1M context; no Fast mode (only Opus 5.5/5/4.8 have it) | Yes | Added | https://platform.claude.com/docs/en/about-claude/pricing https://platform.claude.com/docs/en/models/overview https://platform.claude.com/docs/en/models/sonnet-5-5/overview | New model, released 2026-09-28, now the current Sonnet-tier flagship (Sonnet 5 demoted to legacy). Pricing numerically identical to claude-sonnet-5. matchPattern mirrors claude-sonnet-5 with a `-5` suffix, verified not to collide with it via the match-pattern tester. |+| Anthropic | claude-sonnet-5-5 | Input $2/MTok, Output $10/MTok; 5m $2.50/MTok, 1h $4/MTok, read $0.10/MTok (0.05x, corrected 2026-10-13; was $0.20/MTok standard 0.1x at launch) | Yes | Flat 1M context; no Fast mode (only Opus 5.5/5/4.8 have it) | Yes | Updated | https://platform.claude.com/docs/en/about-claude/pricing https://platform.claude.com/docs/en/models/overview https://platform.claude.com/docs/en/models/sonnet-5-5/overview | Released 2026-09-28, the current Sonnet-tier flagship (Sonnet 5 demoted to legacy). Input/output/cache-write pricing numerically identical to claude-sonnet-5; cache-read price was corrected 2026-10-13 from the standard 0.1x multiplier to a non-standard 0.05x multiplier per the pricing page's cache-hits footnote, which now lists Sonnet 5.5 alongside Opus 5.5 (both 0.05x) separately from the 0.025x Fable 5.1/Mythos 5.1 rate. AWS Bedrock's pricing page independently corroborates a "Starting October 7, 2026" 50%-off cache-read change for this model. matchPattern mirrors claude-sonnet-5 with a `-5` suffix, verified not to collide with it via the match-pattern tester. || Anthropic | claude-sonnet-4-6 | Input $3/MTok, Output $15/MTok, 5m $3.75/MTok, 1h $6/MTok, read $0.30/MTok | Yes | Flat 1M context | Yes | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed unchanged. || Anthropic | claude-sonnet-4-5-20250929 | Input $3/MTok, Output $15/MTok, 5m $3.75/MTok, 1h $6/MTok, read $0.30/MTok | Yes | No large-context tier (200k hard cap) | Yes | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed unchanged. || Anthropic | claude-sonnet-4-20250514 | Input $3/MTok, Output $15/MTok, 5m $3.75/MTok, 1h $6/MTok, read $0.30/MTok | Yes | Retired except Bedrock/Google Cloud β no tiering | Not applicable | None | https://platform.claude.com/docs/en/about-claude/pricing | Re-confirmed present on current page's main table, unchanged. |diff --git a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.mdindex 8a69fb570f..c0ff4a86ca 100644--- a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md+++ b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md@@ -1176,6 +1176,53 @@ json_literal = "<price_per_mtok>e-6"`types.ts` is not one of the selectable arrays the audit may edit; report anewly released Jev version that should become selectable as unresolved.+- **Claude Sonnet 5.5 cache-read price corrected to 0.05x (found October 13 2026)** β+ Anthropic's pricing page cache-hits footnote changed from listing only Claude Opus 5.5+ (and separately Fable 5.1/Mythos 5.1) at a non-standard multiplier to now also including+ Claude Sonnet 5.5: "Cache hits and refreshes on Claude Opus 5.5 and Claude Sonnet 5.5 are+ priced at 0.05x the base input price." This lowers `claude-sonnet-5-5`'s cache-read price+ from $0.20/MTok (the standard 0.1x multiplier it launched with on 2026-09-28, see the+ September 29 2026 entry above) to $0.10/MTok. Input ($2/MTok), output ($10/MTok), and both+ cache-write rates (5m $2.50/MTok, 1h $4/MTok) are unchanged. AWS Bedrock's pricing page+ independently corroborates this with its own footnote: "Starting October 7, 2026, cache+ read pricing for Claude Sonnet 5.5 is reduced by 50% from the prior rate" β consistent+ with $0.20 to $0.10 β but a WebFetch of the Bedrock page could not extract the actual+ numeric pricing row for this or any other Claude model on this run (see the next entry);+ the correction was applied from the direct Anthropic pricing page alone, which is+ authoritative for all Anthropic-operated and partner platforms using the same model ID.+ Updated `cache_read_input_tokens`/`input_cached_tokens`/`input_cache_read` only; no other+ fields changed. Official sources: `https://platform.claude.com/docs/en/about-claude/pricing`.+- **AWS Bedrock pricing page: WebFetch cannot reliably extract per-model numeric rows+ (confirmed October 13 2026)** β Multiple independently-worded WebFetch requests against+ `https://aws.amazon.com/bedrock/pricing/` this run, including ones asking only for a single+ named model's input/output/cache prices, all returned prose acknowledging that a pricing+ section exists but stating the actual numeric table rows were "not included in the provided+ excerpt" or "not fully displayed." This is a harder failure than the Gemini free/paid+ column-collapse issue (that one still returns numbers, just sometimes the wrong column);+ here no numbers come through for Claude models at all, even though the page's surrounding+ prose (e.g. a cache-read-price-change footnote) is readable. Treat any Bedrock pricing claim+ that lacks an actual quoted number as unconfirmed; do not use it alone to change a price,+ but a qualitative claim (e.g. "price X changed by Y%") can still corroborate a figure+ obtained from another official source, as happened with the Claude Sonnet 5.5 correction+ above. Re-attempt a full tabular fetch in a future audit if a Bedrock-only price (such as+ the Public Extended Access SKU) needs independent reconfirmation.+- **Gemini 3.x pricing-page single-model fetch gap for gemini-3.5-flash and+ gemini-3.7-flash (found October 13 2026)** β Targeted WebFetch requests against+ `https://ai.google.dev/gemini-api/docs/pricing` asking only about `gemini-3.5-flash` (the+ non-Lite sibling) or `gemini-3.7-flash` each returned "model not found on this page," even+ though broader fetches of the same page in the same run returned data for+ `gemini-3.8-flash`, `gemini-3.6-flash`, and `gemini-3.5-flash-lite`, and the Gemini models+ catalog (`ai.google.dev/gemini-api/docs/models`) lists both as active (not labeled "Shut+ down") in the same run. This is a different failure mode from the known free/paid+ column-collapse issue: the row appears to be silently dropped rather than+ misattributed. Do not treat a "model not found" result for these two IDs as evidence of+ removal β cross-check the models catalog's shutdown labels first, and prefer a fetch that+ asks for every row on the page at once (risking column-collapse, correctable with a+ Free/Paid-explicit follow-up) over a single-model-targeted fetch on this specific page+ (risking the row being dropped entirely). `gemini-3.5-flash` and `gemini-3.7-flash` prices+ were left unchanged this run, carried forward from their last confirmed values+ ($1.50/$9.00/$0.15 and $0.75/$3.75/$0.075 respectively).+## Provider Usage KeysUse [provider-usage-key-matrix.md](provider-usage-key-matrix.md) as the singlediff --git a/worker/src/constants/default-model-prices.json b/worker/src/constants/default-model-prices.jsonindex cecb2bbc5d..a3036905be 100644--- a/worker/src/constants/default-model-prices.json+++ b/worker/src/constants/default-model-prices.json@@ -4685,7 +4685,7 @@"modelName": "claude-sonnet-5-5","matchPattern": "(?i)^((anthropic\\/)?claude-sonnet-5-5|(eu\\.|us\\.|apac\\.|au\\.|jp\\.|global\\.)?anthropic\\.claude-sonnet-5-5(-v1(:0)?)?)$","createdAt": "2026-09-29T00:00:00.000Z",- "updatedAt": "2026-09-29T00:00:00.000Z",+ "updatedAt": "2026-10-13T00:00:00.000Z","tokenizerConfig": null,"tokenizerId": "claude","pricingTiers": [@@ -4704,9 +4704,9 @@"input_cache_creation": 2.5e-6,"input_cache_creation_5m": 2.5e-6,"input_cache_creation_1h": 4e-6,- "cache_read_input_tokens": 2e-7,- "input_cached_tokens": 2e-7,- "input_cache_read": 2e-7+ "cache_read_input_tokens": 0.1e-6,+ "input_cached_tokens": 0.1e-6,+ "input_cache_read": 0.1e-6}}]