Daily vendor refresh (2026-06-09): OpenAI nano + Mistral Large 3 pricing

research/research_report_2026-06-09.md

@@ -0,0 +1,136 @@
+## 0. Diff & Freshness
+
+- **Report date**: 2026-06-09
+- **Prompt version**: v1
+- **Prior catalog**: `ai-context-api` `main` at commit `a1d2109` (2026-06-03).
+
+### Delta (verified)
+
+- **Google (Gemini)**: Gemini Developer API pricing page currently shows token pricing as “Free of charge” for Gemini 2.5 Pro / 2.5 Flash / 2.5 Flash-Lite and many Gemini 3.x variants; page does not show the previously tracked paid rates for 2.5 Pro/Flash. This is likely a product change or a rendering/segment issue; treat as **⚠ disputed/unclear** until corroborated elsewhere. [1]
+- **OpenAI**: OpenAI pricing page lists `gpt-5.4-nano` at $0.20 input / $1.25 output per 1M tokens, and also lists `gpt-5.5-pro` and `gpt-5.4-pro` tiers. Dataset currently tracks `gpt-5-4-nano` but at placeholders, and does not include `gpt-5.5-pro` / `gpt-5.4-pro`. [2]
+- **DeepSeek**: DeepSeek pricing page documents that `deepseek-chat` and `deepseek-reasoner` will be deprecated on 2026-07-24 15:59 UTC, mapping to non-thinking and thinking modes of `deepseek-v4-flash`. Dataset does not currently record these deprecated aliases. [3]
+- **Mistral**: Official pricing page lists “Mistral Large 3” at $0.5 input / $1.5 output per 1M tokens, not $2/$6 currently in `main`. (There is an open PR #9 that already includes this fix, but it is not merged into `main` yet.) [4]
+
+No other provider changes were found from today’s primary pricing pages.
+
+---
+
+# 1. Models catalog (`Model`)
+
+## Anthropic
+
+No changes found vs current catalog.
+
+## OpenAI
+
+No verified changes to model *catalog* (IDs in dataset) from the pricing page alone; however, the pricing page includes additional SKUs (`gpt-5.5-pro`, `gpt-5.4-pro`) that the dataset does not track. [2]
+
+## Google
+
+Gemini pricing page appears to show “Free of charge” for many models; no reliable paid-tier numbers were visible today, so model catalog changes are held back. [1]
+
+## Meta
+
+No changes (insufficient primary source access for Llama model metadata today).
+
+## Mistral
+
+No model list changes; pricing-only discrepancy captured in §2. [4]
+
+## DeepSeek
+
+No new model IDs; deprecation note for legacy aliases in §2. [3]
+
+## xAI
+
+No model list changes.
+
+## Cohere
+
+No model list changes.
+
+**References (Models)**
+
+[1] https://ai.google.dev/gemini-api/docs/pricing
+
+[2] https://platform.openai.com/docs/pricing
+
+[3] https://api-docs.deepseek.com/quick_start/pricing
+
+[4] https://mistral.ai/pricing
+
+---
+
+# 2. Pricing (`Pricing`)
+
+## Confirmed pricing deltas
+
+### Mistral
+
+- `mistral-large-3`: official list rate shown as **$0.50 input / $1.50 output per 1M tokens**, vs **$2.00 / $6.00** in `main`. [1]
+
+### OpenAI
+
+- `gpt-5.4-nano`: OpenAI pricing table shows **$0.20 input / $0.02 cached input / $1.25 output per 1M tokens**; dataset currently has placeholder values for nano. [2]
+
+## Non-deltas (re-verified)
+
+### Anthropic
+
+Anthropic pricing table still lists:
+- Claude Opus 4.8 and 4.7: $5 input / $0.50 cache hit / $25 output, and batch input $2.50 / batch output $12.50. [3]
+- Claude Sonnet 4.6: $3 input / $0.30 cache hit / $15 output, batch input $1.50 / batch output $7.50. [3]
+- Claude Haiku 4.5: $1 input / $0.10 cache hit / $5 output, batch input $0.50 / batch output $2.50. [3]
+
+### DeepSeek
+
+DeepSeek pricing table still lists:
+- `deepseek-v4-flash`: $0.14 input (cache miss) / $0.0028 input (cache hit) / $0.28 output per 1M tokens. [4]
+- `deepseek-v4-pro`: $0.435 input / $0.87 output per 1M tokens. [4]
+Also notes `deepseek-chat` and `deepseek-reasoner` will be deprecated on 2026/07/24 15:59 UTC. [4]
+
+### Google (Gemini)
+
+Gemini pricing page did not show consistent paid-tier token prices today (most rows show “Free of charge”). Hold changes until a stable paid-tier table is accessible. [5]
+
+### Cohere
+
+Cohere pricing page still states “Command R+ 04-2024 pricing is $3.00 / $15.00” and “Command R+ 08-2024 pricing is $2.50 / $10.00” (existing customers FAQ). [6]
+
+### xAI
+
+xAI models page still lists Grok 4.3 at $1.25 input / $2.50 output per 1M tokens (1M context). [7]
+
+**References (Pricing)**
+
+[1] https://mistral.ai/pricing
+
+[2] https://platform.openai.com/docs/pricing
+
+[3] https://docs.anthropic.com/en/docs/about-claude/pricing
+
+[4] https://api-docs.deepseek.com/quick_start/pricing
+
+[5] https://ai.google.dev/gemini-api/docs/pricing
+
+[6] https://cohere.com/pricing
+
+[7] https://docs.x.ai/docs/models
+
+---
+
+# 3. Benchmarks (`Benchmark`)
+
+No benchmark refresh performed today (no new independent leaderboard snapshot collected in this run).
+
+---
+
+# 4. Use-cases (`UseCase`)
+
+No changes.
+
+---
+
+# 5. Historic-usage (`HistoricUsageCase`)
+
+No changes.

research/research_report_2026-06-11.md

@@ -0,0 +1,79 @@
+## 0. Diff & Freshness
+
+- **Report date**: 2026-06-11
+- **Prompt version**: v1
+- **Branch / PR**: `data-refresh-2026-06-09` (PR #10).
+- **Prior catalog**: `ai-context-api` `main` at commit `a1d2109` (2026-06-03), plus the data edits already staged on this branch (`pricing.ts`).
+- **Purpose**: Re-verify the pricing deltas already carried by PR #10 against today's primary vendor pages, and record OpenAI mapping items deliberately held back. No new data rows were added in this run beyond what PR #10 already contains.
+
+### Re-verification summary (2026-06-11)
+
+- **Mistral — confirmed, no change needed.** `mistral.ai/pricing` continues to list **Mistral Large 3** at **$0.50 input / $1.50 output per 1M tokens** with a stated **50% batch discount** (→ $0.25 batch input). This matches the values already on the branch for `mistral-large-3` (`inputPerMTokUsd: 0.5`, `outputPerMTokUsd: 1.5`, `batchInputPerMTokUsd: 0.25`). "Mistral Large 3" maps to the existing dataset ID `mistral-large-3`. [1]
+- **OpenAI — confirmed for the tracked SKU, no change needed.** The OpenAI API pricing page (`platform.openai.com/docs/pricing`, now serving from `developers.openai.com/api/docs/pricing`) lists **gpt-5.4-nano** at **$0.20 input / $0.02 cached input / $1.25 output per 1M tokens**. This matches the row already on the branch for the existing dataset ID `gpt-5-4-nano`. The dataset's dash-normalized ID `gpt-5-4-nano` corresponds to OpenAI's `gpt-5.4-nano`. [2]
+
+### OpenAI items held back (mapping ambiguity)
+
+The same OpenAI pricing table lists two additional SKUs that the dataset does **not** track:
+
+- **gpt-5.5-pro**: $30.00 input / $180.00 output per 1M tokens. [2]
+- **gpt-5.4-pro**: $30.00 input / $180.00 output per 1M tokens. [2]
+
+These are **held back** and intentionally not added, because:
+
+1. There is no corresponding model entry in `src/data/models.ts` (the dataset tracks `gpt-5-5`, `gpt-5-4`, `gpt-5-4-mini`, `gpt-5-4-nano`). A `Pricing` row references a `modelId`; adding a price row without a matching `Model` would introduce a dangling ID.
+2. The pricing page alone does not establish stable model metadata (context window, max output, modalities, release date) needed for a conservative `Model` row.
+3. Adding "pro" tiers is a catalog expansion, not a pricing correction, and is outside the conservative scope of this refresh.
+
+Recommendation: defer `gpt-5.5-pro` / `gpt-5.4-pro` to a dedicated catalog-expansion change once primary model docs (not just the pricing table) confirm metadata and a stable ID convention.
+
+### Other providers (re-verified, no change)
+
+- **Cohere**: Command R+ SKU reconciliation (Command R+ 08-2024 at $2.50/$10.00) is handled separately in PR #9 and is **not** touched here to avoid duplicate edits. PR #10 leaves `command-r-plus-2025` ($3.00/$15.00) unchanged. [3]
+- **xAI**: `docs.x.ai/docs/models` still lists Grok 4.3 at $1.25 input / $2.50 output. Grok Build 0.1 is **held back** — it does not yet fit the repo's locked canonical provider/model strategy and lacks the required catalog metadata. [4]
+- **Google (Gemini)**: pricing page continues to render inconsistent paid-tier numbers ("Free of charge" rows). No change; held until a stable paid-tier table is accessible.
+- **DeepSeek / Anthropic / Meta**: no verified deltas vs. current dataset today.
+
+---
+
+# 1. Models catalog (`Model`)
+
+No model catalog changes in this run. Held-back OpenAI "pro" SKUs documented in §0.
+
+---
+
+# 2. Pricing (`Pricing`)
+
+No new pricing rows added on 2026-06-11. The two pricing edits already present on branch `data-refresh-2026-06-09` were re-verified against primary sources today and remain correct:
+
+- `gpt-5-4-nano`: $0.20 input / $0.02 cached / $1.25 output per 1M tokens. [2]
+- `mistral-large-3`: $0.50 input / $1.50 output per 1M tokens; batch input $0.25 (50% discount). [1]
+
+---
+
+# 3. Benchmarks (`Benchmark`)
+
+No benchmark refresh in this run.
+
+---
+
+# 4. Use-cases (`UseCase`)
+
+No changes.
+
+---
+
+# 5. Historic-usage (`HistoricUsageCase`)
+
+No changes.
+
+---
+
+**References**
+
+[1] https://mistral.ai/pricing
+
+[2] https://platform.openai.com/docs/pricing (redirects to https://developers.openai.com/api/docs/pricing)
+
+[3] https://cohere.com/pricing
+
+[4] https://docs.x.ai/docs/models

src/data/pricing.ts

@@ -55,6 +55,13 @@ export const pricing: Pricing[] = [
     batchInputPerMTokUsd: 0.375,
     notes: "OpenAI API pricing table (short context).",
   },
+  {
+    modelId: "gpt-5-4-nano",
+    inputPerMTokUsd: 0.20,
+    outputPerMTokUsd: 1.25,
+    cachedInputPerMTokUsd: 0.02,
+    notes: "OpenAI API pricing table (short context).",
+  },
   {
     modelId: "gemini-2-5-pro",
     inputPerMTokUsd: 1.25,
@@ -95,9 +102,10 @@ export const pricing: Pricing[] = [
   },
   {
     modelId: "mistral-large-3",
-    inputPerMTokUsd: 2.0,
-    outputPerMTokUsd: 6.0,
-    batchInputPerMTokUsd: 1.0,
+    inputPerMTokUsd: 0.5,
+    outputPerMTokUsd: 1.5,
+    batchInputPerMTokUsd: 0.25,
+    notes: "Mistral pricing page lists $0.50/$1.50 per 1M tokens; batch processing gets a 50% discount.",
   },
     {
     modelId: "deepseek-v4-flash",