Merge pull request #843 from cgilly2fast/dev

chore: clean up firmware available models

providers/firmware/models/claude-haiku-4-5.toml

@@ -3,15 +3,17 @@ family = "claude-haiku"
 release_date = "2025-10-15"
 last_updated = "2025-10-15"
 attachment = true
-reasoning = true
+reasoning = false
 temperature = true
 tool_call = true
 knowledge = "2025-02-28"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25
 
 [limit]
 context = 200_000

providers/firmware/models/claude-opus-4-5.toml

@@ -3,15 +3,17 @@ family = "claude-opus"
 release_date = "2025-11-24"
 last_updated = "2025-11-24"
 attachment = true
-reasoning = true
+reasoning = false
 temperature = true
 tool_call = true
 knowledge = "2025-03-31"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 200_000

providers/firmware/models/claude-opus-4-6.toml

@@ -3,15 +3,17 @@ family = "claude-opus"
 release_date = "2026-02-05"
 last_updated = "2026-02-05"
 attachment = true
-reasoning = true
+reasoning = false
 temperature = true
 tool_call = true
 knowledge = "2025-05-31"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
 
 [limit]
 context = 200_000

providers/firmware/models/claude-sonnet-4-5.toml

@@ -3,15 +3,17 @@ family = "claude-sonnet"
 release_date = "2025-09-29"
 last_updated = "2025-09-29"
 attachment = true
-reasoning = true
+reasoning = false
 temperature = true
 tool_call = true
 knowledge = "2025-07-31"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [limit]
 context = 200_000

providers/firmware/models/deepseek-r1.toml

@@ -1,9 +1,9 @@
-name = "DeepSeek Reasoner"
+name = "DeepSeek R1"
 family = "deepseek-thinking"
 release_date = "2025-01-20"
 last_updated = "2025-09-29"
 attachment = true
-reasoning = true
+reasoning = false
 temperature = true
 knowledge = "2024-07"
 tool_call = true
@@ -13,8 +13,8 @@ open_weights = false
 field = "reasoning_content"
 
 [cost]
-input = 0
-output = 0
+input = 1.35
+output = 5.4
 
 [limit]
 context = 128_000

providers/firmware/models/deepseek-v3.toml

@@ -1,4 +1,4 @@
-name = "DeepSeek Chat"
+name = "DeepSeek v3.2"
 family = "deepseek"
 release_date = "2024-12-26"
 last_updated = "2025-09-29"
@@ -10,8 +10,8 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 0.58
+output = 1.68
 
 [limit]
 context = 128_000

providers/firmware/models/gemini-2.5-flash.toml

@@ -10,8 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 0.3
+output = 2.5
+cache_read = 0.075
 
 [limit]
 context = 1_048_576

providers/firmware/models/gemini-2.5-pro.toml

@@ -10,8 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 1.25
+output = 10
+cache_read = 0.31
 
 [limit]
 context = 1_048_576

providers/firmware/models/gemini-3-flash-preview.toml

@@ -11,8 +11,9 @@ knowledge = "2025-01"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 0.5
+output = 3
+cache_read = 0.05
 
 [limit]
 context = 1_048_576

providers/firmware/models/gemini-3-pro-preview.toml

@@ -11,8 +11,9 @@ knowledge = "2025-01"
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 2
+output = 12
+cache_read = 0.2
 
 [limit]
 context = 1_000_000

providers/firmware/models/gpt-4o.toml

@@ -11,8 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 2.5
+output = 10
+cache_read = 1.25
 
 [limit]
 context = 128_000

providers/firmware/models/gpt-5-mini.toml

@@ -11,9 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
-cache_read = 0
+input = 0.25
+output = 2
+cache_read = 0.03
 
 [limit]
 context = 400_000

providers/firmware/models/gpt-5-nano.toml

@@ -11,9 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
-cache_read = 0
+input = 0.05
+output = 0.4
+cache_read = 0.01
 
 [limit]
 context = 400_000

providers/firmware/models/gpt-5.2.toml

@@ -10,8 +10,9 @@ tool_call = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 1.75
+output = 14
+cache_read = 0.175
 
 [limit]
 context = 400_000

providers/firmware/models/gpt-5.toml

@@ -11,8 +11,9 @@ structured_output = true
 open_weights = false
 
 [cost]
-input = 0
-output = 0
+input = 1.25
+output = 10
+cache_read = 0.13
 
 [limit]
 context = 400_000

providers/firmware/models/gpt-oss-120b.toml

@@ -10,8 +10,8 @@ structured_output = true
 open_weights = true
 
 [cost]
-input = 0
-output = 0
+input = 0.15
+output = 0.6
 
 [limit]
 context = 131_072

providers/firmware/models/gpt-oss-20b.toml

@@ -1,20 +1,21 @@
-name = "Z.AI GLM-4.7"
+name = "GPT OSS 120B"
+family = "gpt-oss"
 release_date = "1970-01-01"
 last_updated = "1970-01-01"
 attachment = false
-reasoning = false
+reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
 open_weights = true
 
 [cost]
-input = 0
-output = 0
+input = 0.07
+output = 0.2
 
 [limit]
 context = 131_072
-output = 40_000
+output = 32_768
 
 [modalities]
 input = ["text"]

providers/firmware/models/grok-4-fast-non-reasoning.toml

@@ -1,22 +0,0 @@
-name = "Grok 4 Fast (Non-Reasoning)"
-family = "grok"
-release_date = "2025-09-19"
-last_updated = "2025-09-19"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2025-07"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 2_000_000
-output = 30_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/firmware/models/grok-4-fast-reasoning.toml

@@ -1,22 +0,0 @@
-name = "Grok 4 Fast (Reasoning)"
-family = "grok"
-release_date = "2025-09-19"
-last_updated = "2025-09-19"
-attachment = true
-reasoning = true
-temperature = true
-knowledge = "2025-07"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 2_000_000
-output = 30_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/firmware/models/grok-code-fast-1.toml

@@ -1,22 +0,0 @@
-name = "Grok Code Fast 1"
-family = "grok"
-release_date = "2025-08-28"
-last_updated = "2025-08-28"
-attachment = false
-reasoning = true
-temperature = true
-knowledge = "2023-10"
-tool_call = true
-open_weights = false
-
-[cost]
-input = 0
-output = 0
-
-[limit]
-context = 256_000
-output = 10_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/firmware/models/kimi-k2-thinking-turbo.toml

@@ -1,21 +0,0 @@
-name = "Kimi K2 Thinking Turbo"
-release_date = "1970-01-01"
-last_updated = "1970-01-01"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 1.15
-output = 8.0
-
-[limit]
-context = 262_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/firmware/models/kimi-k2-thinking.toml

@@ -1,21 +0,0 @@
-name = "Kimi K2 Thinking"
-release_date = "1970-01-01"
-last_updated = "1970-01-01"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.6
-output = 2.5
-
-[limit]
-context = 262_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/firmware/models/kimi-k2.5.toml

@@ -1,21 +0,0 @@
-name = "Kimi-K2.5"
-release_date = "1970-01-01"
-last_updated = "1970-01-01"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[cost]
-input = 0.1
-output = 3.0
-
-[limit]
-context = 262_000
-output = 128_000
-
-[modalities]
-input = ["text"]
-output = ["text"]