@oh-my-pi/pi-catalog 18.13.0 → 18.13.1

package/CHANGELOG.md

@@ -2,6 +2,13 @@
 
 ## [Unreleased]
 
+## [18.13.1] - 2026-10-27
+
+### Added
+
+- Added GPT-6.1 Terra: `openai/gpt-6.1-terra` ships bundled ($1/$5, $0.05 cached, >272K long-context tier, 1.05M window), and Codex-discovered `gpt-6.1-terra`/`-wm` get API-equivalent pricing, the 922K extended-context maximum, and routing to Plus-or-higher accounts.
+- Added Kimi K3.1 Pro and K3.1 Pro Highspeed support on Kimi Code (`k3.1-pro`, `k3.1-pro-highspeed`) and the Moonshot API (`kimi-k3.1-pro`, `kimi-k3.1-pro-highspeed`): official pricing ($2/$10 and $4/$20 per MTok), a 131K output cap, and K3's reasoning and wire contract ([#16901](https://github.com/can1357/oh-my-pi/pull/16901) by [@eggpeat](https://github.com/eggpeat))
+
 ## [18.13.0] - 2026-10-26
 
 ### Breaking Changes

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@oh-my-pi/pi-catalog",
-	"version": "18.13.0",
+	"version": "18.13.1",
 	"description": "Model catalog for omp: bundled model database, provider discovery descriptors, model identity, classification, and equivalence",
 	"keywords": [
 		"ai",
@@ -93,12 +93,12 @@
 		"gen:proto": "bun scripts/generate-protocols.ts"
 	},
 	"dependencies": {
-		"@oh-my-pi/omptype": "18.13.0",
-		"@oh-my-pi/pi-utils": "18.13.0"
+		"@oh-my-pi/omptype": "18.13.1",
+		"@oh-my-pi/pi-utils": "18.13.1"
 	},
 	"devDependencies": {
 		"@bgotink/kdl": "0.4.0",
-		"@oh-my-pi/pi-ai": "18.13.0",
+		"@oh-my-pi/pi-ai": "18.13.1",
 		"@types/bun": "^1.3.14"
 	},
 	"engines": {

package/src/compat/rules/providers/kimi-code.kdl

@@ -55,6 +55,22 @@ provider "kimi-code" {
 			cache-write 3.0
 		}
 	}
+	models "k3.1-pro" {
+		cost-fallback {
+			input 2.0
+			output 10.0
+			cache-read 0.2
+			cache-write 2.0
+		}
+	}
+	models "k3.1-pro-highspeed" {
+		cost-fallback {
+			input 4.0
+			output 20.0
+			cache-read 0.2
+			cache-write 4.0
+		}
+	}
 	models "kimi-for-coding" {
 		cost-fallback {
 			input 0.95

package/src/compat/rules/providers/moonshot.kdl

@@ -14,6 +14,12 @@ provider "moonshot" {
 		thinking-format "zai"
 		family "k3" {
 			thinking-format "openai"
+			// K3 and K3.1 Pro default `max_completion_tokens` to 131072 and
+			// accept up to the whole 1M window; the shared catalog reports that
+			// ceiling as the output limit, leaving no room for input. Pin the
+			// documented default, as `model-limits` does for /models rows.
+			// https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart
+			max-tokens-ceiling 131072
 		}
 		thinking-mode "effort"
 	}
@@ -44,6 +50,25 @@ provider "moonshot" {
 			cache-write 3.0
 		}
 	}
+	// Official K3.1 Pro rates. The discovery mapper seeds K3 list pricing on
+	// every $0 K3-family row, so these must patch rather than fall back.
+	// https://platform.kimi.ai/docs/pricing/chat
+	models "kimi-k3.1-pro" {
+		cost-patch {
+			input 2.0
+			output 10.0
+			cache-read 0.2
+			cache-write 2.0
+		}
+	}
+	models "kimi-k3.1-pro-highspeed" {
+		cost-patch {
+			input 4.0
+			output 20.0
+			cache-read 0.2
+			cache-write 4.0
+		}
+	}
 	// The /models roster carries no modality metadata for the named vision/vl
 	// SKUs; the bounded id token is the census signal.
 	models token="vision" {

package/src/compat/rules/providers/openai-codex.kdl

@@ -140,6 +140,19 @@ provider "openai-codex" {
 			ultrafast 8
 		}
 	}
+	// GPT-6.1 Terra lists at half GPT-6.1 Sol: $1/$5 with $0.05 cached input
+	// (OpenAI API docs, 2026-10-26). OpenAI documents 922K max input (1.05M
+	// total); the curated maximum applies only with extended context on, and a
+	// higher live maximum still wins. Codex ships no Ultrafast tier for it.
+	models "gpt-6.1-terra" "gpt-6.1-terra-wm" {
+		cost-patch {
+			input 1.0
+			output 5.0
+			cache-read 0.05
+			cache-write 0
+		}
+		max-context-window 922000
+	}
 	models "gpt-6-luna" "gpt-6-luna-wm" {
 		cost-patch {
 			input 0.1

package/src/compat/rules/providers/openai.kdl

@@ -204,9 +204,10 @@ provider "openai" {
 			cache-write 25.0
 		}
 	}
-	// GPT-6 Sol and Luna use the same >272K whole-request multipliers as Astra.
-	// Derive the tier from each API card so upstream price changes carry through.
-	models "gpt-6-sol" "gpt-6-luna" {
+	// GPT-6 Sol, GPT-6 Luna, and GPT-6.1 Terra use the same >272K whole-request
+	// multipliers as Astra. Derive the tier from each API card so upstream price
+	// changes carry through.
+	models "gpt-6-sol" "gpt-6-luna" "gpt-6.1-terra" {
 		long-context-cost {
 			input-threshold 272000
 			multiplier 2.0

package/src/compat/rules/runtime/behavior.kdl

@@ -114,11 +114,14 @@ behavior {
 			"aster-07-15" "gpt-image-2.5-sunburst" "gpt-image-2.5-flare" "gemini-3-pro-image-preview"
 	}
 	// Subscription eligibility for model-aware Codex credential routing.
+	// GPT-6.1 Terra launched for Plus, Pro, Enterprise, and Edu, not Free/Go
+	// (https://openai.com/index/introducing-gpt-6-1-terra/, 2026-10-26).
 	plan-requirement provider="openai-codex" {
 		tier "pro" substring="-spark"
 		tier "paid" exact="gpt-5.6" exact="gpt-5.6-sol" exact="gpt-5.6-sol-pro" \
 			exact="gpt-5.6-luna" exact="gpt-5.6-luna-pro" glob="*/gpt-5.6" glob="*/gpt-5.6-sol" \
-			glob="*/gpt-5.6-sol-pro" glob="*/gpt-5.6-luna" glob="*/gpt-5.6-luna-pro"
+			glob="*/gpt-5.6-sol-pro" glob="*/gpt-5.6-luna" glob="*/gpt-5.6-luna-pro" \
+			glob="gpt-6.1-terra*" glob="*/gpt-6.1-terra*"
 	}
 
 	// Z.AI and Zhipu emit timezone-naive account-reset wall clocks in Beijing
@@ -335,11 +338,14 @@ behavior {
 	}
 	// Kimi Code's /coding/v1/models envelope carries no output-limit field;
 	// these are the documented per-family caps (models.dev `kimi-for-coding`
-	// and `moonshotai` kimi-k3 entries, #6711). Legacy K2 rows keep their
+	// and `moonshotai` kimi-k3 entries, #6711; K3.1 Pro's documented
+	// `max_completion_tokens` default). Legacy K2 rows keep their
 	// discovery-reported values.
 	model-limits provider="kimi-code" {
 		limits "k3" max-tokens=131072
 		limits "k3-256k" max-tokens=131072
+		limits "k3.1-pro" max-tokens=131072
+		limits "k3.1-pro-highspeed" max-tokens=131072
 		limits "kimi-for-coding" max-tokens=32768
 		limits "kimi-for-coding-highspeed" max-tokens=32768
 	}

package/src/compat/rules/taxonomy/kimi.kdl

@@ -19,6 +19,13 @@ class "kimi" {
     family "k3" glob="k3"
     family "k3" glob="k3-256k"
 
+    // K3.1 Pro is a post-trained K3 with K3's full request contract (Moonshot
+    // documents identical parameter constraints), so it keeps family `k3` and
+    // inherits every K3 lineage rule. The revision gives it its own weights
+    // identity, so K3 reasoning never replays natively into K3.1 Pro. Kimi
+    // Code's bare `k3.1-pro*` ids reach no class matcher without these.
+    override id="kimi-k3-1-pro" glob="*k3.1-pro" class="kimi" family="k3" revision="3.1" rationale="K3.1 Pro is a post-trained K3 sharing K3's parameter contract" provenance="https://platform.kimi.ai/docs/api/models-overview"
+    override id="kimi-k3-1-pro-highspeed" glob="*k3.1-pro-highspeed" class="kimi" family="k3" revision="3.1" rationale="K3.1 Pro Highspeed serves K3.1 Pro faster with identical parameter constraints" provenance="https://platform.kimi.ai/docs/api/models-overview"
     override id="kilo-kimi-k2-thinking-product" provider="kilo" model="kimi-k2-thinking" logical="moonshotai/kimi-k2-thinking" class="kimi" rationale="Oracle exposes this independently priced product rather than an effort route" provenance="fixtures/llm-oracle/catalog/models.normalized.json"
     override id="openrouter-kimi-k2-thinking-product" provider="openrouter" model="kimi-k2-thinking" logical="moonshotai/kimi-k2-thinking" class="kimi" rationale="Oracle exposes this independently priced product rather than an effort route" provenance="fixtures/llm-oracle/catalog/models.normalized.json"
     override id="vercel-kimi-k2-thinking-product" provider="vercel-ai-gateway" model="kimi-k2-thinking" logical="moonshotai/kimi-k2-thinking" class="kimi" rationale="Oracle exposes this independently priced product rather than an effort route" provenance="fixtures/llm-oracle/catalog/models.normalized.json"