groq.go 2.5 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889
  1. package models
  2. const (
  3. ProviderGROQ ModelProvider = "groq"
  4. // GROQ
  5. QWENQwq ModelID = "qwen-qwq"
  6. // GROQ preview models
  7. Llama4Scout ModelID = "meta-llama/llama-4-scout-17b-16e-instruct"
  8. Llama4Maverick ModelID = "meta-llama/llama-4-maverick-17b-128e-instruct"
  9. Llama3_3_70BVersatile ModelID = "llama-3.3-70b-versatile"
  10. DeepseekR1DistillLlama70b ModelID = "deepseek-r1-distill-llama-70b"
  11. )
  12. var GroqModels = map[ModelID]Model{
  13. //
  14. // GROQ
  15. QWENQwq: {
  16. ID: QWENQwq,
  17. Name: "Qwen Qwq",
  18. Provider: ProviderGROQ,
  19. APIModel: "qwen-qwq-32b",
  20. CostPer1MIn: 0.29,
  21. CostPer1MInCached: 0.275,
  22. CostPer1MOutCached: 0.0,
  23. CostPer1MOut: 0.39,
  24. ContextWindow: 128_000,
  25. DefaultMaxTokens: 50000,
  26. // for some reason, the groq api doesn't like the reasoningEffort parameter
  27. CanReason: false,
  28. SupportsAttachments: false,
  29. },
  30. Llama4Scout: {
  31. ID: Llama4Scout,
  32. Name: "Llama4Scout",
  33. Provider: ProviderGROQ,
  34. APIModel: "meta-llama/llama-4-scout-17b-16e-instruct",
  35. CostPer1MIn: 0.11,
  36. CostPer1MInCached: 0,
  37. CostPer1MOutCached: 0,
  38. CostPer1MOut: 0.34,
  39. DefaultMaxTokens: 8192,
  40. ContextWindow: 128_000, // 10M when?
  41. SupportsAttachments: true,
  42. },
  43. Llama4Maverick: {
  44. ID: Llama4Maverick,
  45. Name: "Llama4Maverick",
  46. Provider: ProviderGROQ,
  47. APIModel: "meta-llama/llama-4-maverick-17b-128e-instruct",
  48. CostPer1MIn: 0.20,
  49. CostPer1MInCached: 0,
  50. CostPer1MOutCached: 0,
  51. CostPer1MOut: 0.20,
  52. DefaultMaxTokens: 8192,
  53. ContextWindow: 128_000,
  54. SupportsAttachments: true,
  55. },
  56. Llama3_3_70BVersatile: {
  57. ID: Llama3_3_70BVersatile,
  58. Name: "Llama3_3_70BVersatile",
  59. Provider: ProviderGROQ,
  60. APIModel: "llama-3.3-70b-versatile",
  61. CostPer1MIn: 0.59,
  62. CostPer1MInCached: 0,
  63. CostPer1MOutCached: 0,
  64. CostPer1MOut: 0.79,
  65. ContextWindow: 128_000,
  66. SupportsAttachments: false,
  67. },
  68. DeepseekR1DistillLlama70b: {
  69. ID: DeepseekR1DistillLlama70b,
  70. Name: "DeepseekR1DistillLlama70b",
  71. Provider: ProviderGROQ,
  72. APIModel: "deepseek-r1-distill-llama-70b",
  73. CostPer1MIn: 0.75,
  74. CostPer1MInCached: 0,
  75. CostPer1MOutCached: 0,
  76. CostPer1MOut: 0.99,
  77. ContextWindow: 128_000,
  78. CanReason: true,
  79. SupportsAttachments: false,
  80. },
  81. }