chore(deepinfra): add 45 new models and update pricing

providers/deepinfra/models/Gryphe/MythoMax-L2-13b.toml

@@ -0,0 +1,22 @@
+# https://deepinfra.com/Gryphe/MythoMax-L2-13b
+name = "MythoMax L2 13B"
+family = "llama"
+release_date = "2023-11-01"
+last_updated = "2023-11-01"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+open_weights = true
+
+[cost]
+input = 0.40
+output = 0.40
+
+[limit]
+context = 4096
+output = 2048
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/MiniMaxAI/MiniMax-M2.1.toml

@@ -1,3 +1,4 @@
+# https://deepinfra.com/MiniMaxAI/MiniMax-M2.1
 name = "MiniMax M2.1"
 release_date = "2025-12-23"
 last_updated = "2025-12-23"
@@ -11,12 +12,12 @@ knowledge = "2025-06"
 [cost]
 input = 0.28
 output = 1.20
-cached_read = 0.14
+cache_read = 0.14
 
 [limit]
 context = 196_608
 output = 196_608
- 
+
 [modalities]
 input = ["text"]
 output = ["text"]

providers/deepinfra/models/MiniMaxAI/MiniMax-M2.5.toml

@@ -11,10 +11,9 @@ knowledge = "2025-06"
 open_weights = true
 
 [cost]
-input = 0.27
-output = 0.95
+input = 0.15
+output = 1.15
 cache_read = 0.03
-cache_write = 0.375
 
 [limit]
 context = 204_800
@@ -26,3 +25,4 @@ output = ["text"]
 
 [interleaved]
 field = "reasoning_content"
+

providers/deepinfra/models/MiniMaxAI/MiniMax-M2.toml

@@ -1,3 +1,4 @@
+# https://deepinfra.com/MiniMaxAI/MiniMax-M2
 name = "MiniMax M2"
 family = "minimax"
 release_date = "2025-11-13"
@@ -15,7 +16,7 @@ field = "reasoning_content"
 [cost]
 input = 0.254
 output = 1.02
-cached_input = 0.127
+cache_read = 0.127
 
 [limit]
 context = 262_144
@@ -24,3 +25,4 @@ output = 32_768
 [modalities]
 input = ["text"]
 output = ["text"]
+

providers/deepinfra/models/NousResearch/Hermes-3-Llama-3.1-405B.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/NousResearch/Hermes-3-Llama-3.1-405B
+name = "Hermes 3 Llama 3.1 405B"
+family = "hermes"
+release_date = "2024-10-01"
+last_updated = "2024-10-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-06"
+open_weights = true
+
+[cost]
+input = 1.00
+output = 1.00
+
+[limit]
+context = 131_072
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/NousResearch/Hermes-3-Llama-3.1-70B.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/NousResearch/Hermes-3-Llama-3.1-70B
+name = "Hermes 3 Llama 3.1 70B"
+family = "hermes"
+release_date = "2024-10-01"
+last_updated = "2024-10-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-06"
+open_weights = true
+
+[cost]
+input = 0.30
+output = 0.30
+
+[limit]
+context = 131_072
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen2.5-72B-Instruct.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen2.5-72B-Instruct
+name = "Qwen2.5 72B Instruct"
+family = "qwen"
+release_date = "2024-12-01"
+last_updated = "2024-12-01"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2024-06"
+open_weights = true
+
+[cost]
+input = 0.36
+output = 0.40
+
+[limit]
+context = 32_768
+output = 8192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-14B.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-14B
+name = "Qwen3 14B"
+family = "qwen"
+release_date = "2025-04-01"
+last_updated = "2025-04-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-10"
+open_weights = true
+
+[cost]
+input = 0.12
+output = 0.24
+
+[limit]
+context = 40_960
+output = 8192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-235B-A22B-Instruct-2507.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-235B-A22B-Instruct-2507
+name = "Qwen3 235B A22B Instruct 2507"
+family = "qwen"
+release_date = "2025-07-01"
+last_updated = "2025-07-01"
+attachment = true
+reasoning = false
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[cost]
+input = 0.071
+output = 0.10
+
+[limit]
+context = 262_144
+output = 81_920
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-235B-A22B-Thinking-2507.toml

@@ -0,0 +1,24 @@
+# https://deepinfra.com/Qwen/Qwen3-235B-A22B-Thinking-2507
+name = "Qwen3 235B A22B Thinking 2507"
+family = "qwen"
+release_date = "2025-07-01"
+last_updated = "2025-07-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[cost]
+input = 0.23
+output = 2.30
+cache_read = 0.20
+
+[limit]
+context = 262_144
+output = 81_920
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-30B-A3B.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-30B-A3B
+name = "Qwen3 30B A3B"
+family = "qwen"
+release_date = "2025-04-01"
+last_updated = "2025-04-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-10"
+open_weights = true
+
+[cost]
+input = 0.08
+output = 0.28
+
+[limit]
+context = 40_960
+output = 8192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-32B.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-32B
+name = "Qwen3 32B"
+family = "qwen"
+release_date = "2025-04-01"
+last_updated = "2025-04-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2024-10"
+open_weights = true
+
+[cost]
+input = 0.08
+output = 0.28
+
+[limit]
+context = 40_960
+output = 8192
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo.toml

@@ -1,3 +1,4 @@
+# https://deepinfra.com/Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo
 name = "Qwen3 Coder 480B A35B Instruct Turbo"
 family = "qwen"
 release_date = "2025-07-23"
@@ -10,13 +11,14 @@ tool_call = true
 open_weights = true
 
 [cost]
-input = 0.3
-output = 1.2
+input = 0.30
+output = 1.00
+cache_read = 0.10
 
 [limit]
 context = 262_144
 output = 66_536
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-Coder-480B-A35B-Instruct.toml

@@ -1,3 +1,4 @@
+# https://deepinfra.com/Qwen/Qwen3-Coder-480B-A35B-Instruct
 name = "Qwen3 Coder 480B A35B Instruct"
 family = "qwen"
 release_date = "2025-07-23"
@@ -19,4 +20,4 @@ output = 66_536
 
 [modalities]
 input = ["text"]
-output = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-Next-80B-A3B-Instruct.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-Next-80B-A3B-Instruct
+name = "Qwen3 Next 80B A3B Instruct"
+family = "qwen"
+release_date = "2026-03-01"
+last_updated = "2026-03-01"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-01"
+open_weights = true
+
+[cost]
+input = 0.09
+output = 1.10
+
+[limit]
+context = 262_144
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-VL-235B-A22B-Instruct.toml

@@ -0,0 +1,24 @@
+# https://deepinfra.com/Qwen/Qwen3-VL-235B-A22B-Instruct
+name = "Qwen3 VL 235B A22B Instruct"
+family = "qwen"
+release_date = "2025-05-01"
+last_updated = "2025-05-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[cost]
+input = 0.20
+output = 0.88
+cache_read = 0.11
+
+[limit]
+context = 262_144
+output = 81_920
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]

providers/deepinfra/models/Qwen/Qwen3-VL-30B-A3B-Instruct.toml

@@ -0,0 +1,23 @@
+# https://deepinfra.com/Qwen/Qwen3-VL-30B-A3B-Instruct
+name = "Qwen3 VL 30B A3B Instruct"
+family = "qwen"
+release_date = "2025-05-01"
+last_updated = "2025-05-01"
+attachment = true
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+
+[cost]
+input = 0.15
+output = 0.60
+
+[limit]
+context = 262_144
+output = 81_920
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]