smoltalk 0.3.1 → 0.4.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (16) hide show
  1. package/dist/classes/message/AssistantMessage.d.ts +2 -0
  2. package/dist/clients/anthropic.d.ts +7 -0
  3. package/dist/clients/anthropic.js +168 -17
  4. package/dist/clients/google.js +5 -2
  5. package/dist/clients/openai.js +5 -2
  6. package/dist/clients/openaiResponses.js +5 -2
  7. package/dist/model.d.ts +2 -0
  8. package/dist/model.js +38 -5
  9. package/dist/models.d.ts +169 -34
  10. package/dist/models.js +83 -17
  11. package/dist/types/costEstimate.d.ts +2 -0
  12. package/dist/types/costEstimate.js +2 -0
  13. package/dist/types/tokenUsage.d.ts +2 -0
  14. package/dist/types/tokenUsage.js +2 -0
  15. package/dist/types.d.ts +4 -0
  16. package/package.json +1 -1
package/dist/models.d.ts CHANGED
@@ -16,6 +16,7 @@ export type BaseModel = {
16
16
  description?: string;
17
17
  inputTokenCost?: number;
18
18
  cachedInputTokenCost?: number;
19
+ cacheCreationInputTokenCost?: number;
19
20
  outputTokenCost?: number;
20
21
  disabled?: boolean;
21
22
  costUnit?: "tokens" | "characters" | "minutes";
@@ -41,6 +42,16 @@ export type TextModel = BaseModel & {
41
42
  levels?: readonly string[];
42
43
  /** Default reasoning level */
43
44
  defaultLevel?: string;
45
+ /**
46
+ * Anthropic thinking API shape:
47
+ * "adaptive" → `thinking: {type: "adaptive"}` + `output_config: {effort}`
48
+ * (Opus 4.6/4.7/4.8, Sonnet 4.6). Required on 4.7+, where the
49
+ * legacy `enabled` form returns a 400.
50
+ * "budget" → `thinking: {type: "enabled", budget_tokens: N}`
51
+ * (Haiku 4.5 and older models).
52
+ * Omit for non-Anthropic models.
53
+ */
54
+ thinkingStyle?: "adaptive" | "budget";
44
55
  /** Whether reasoning/thinking can be fully disabled */
45
56
  canDisable?: boolean;
46
57
  /** Whether the response includes visible thinking content (thinking blocks/parts) */
@@ -250,7 +261,7 @@ export declare const textModels: readonly [{
250
261
  readonly outputsThinking: false;
251
262
  readonly outputsSignatures: false;
252
263
  };
253
- readonly provider: "openai";
264
+ readonly provider: "openai-responses";
254
265
  }, {
255
266
  readonly type: "text";
256
267
  readonly modelName: "gpt-5-mini";
@@ -268,7 +279,7 @@ export declare const textModels: readonly [{
268
279
  readonly outputsThinking: false;
269
280
  readonly outputsSignatures: false;
270
281
  };
271
- readonly provider: "openai";
282
+ readonly provider: "openai-responses";
272
283
  }, {
273
284
  readonly type: "text";
274
285
  readonly modelName: "gpt-5-nano";
@@ -286,7 +297,7 @@ export declare const textModels: readonly [{
286
297
  readonly outputsThinking: false;
287
298
  readonly outputsSignatures: false;
288
299
  };
289
- readonly provider: "openai";
300
+ readonly provider: "openai-responses";
290
301
  }, {
291
302
  readonly type: "text";
292
303
  readonly modelName: "gpt-5.1";
@@ -303,7 +314,7 @@ export declare const textModels: readonly [{
303
314
  readonly outputsThinking: false;
304
315
  readonly outputsSignatures: false;
305
316
  };
306
- readonly provider: "openai";
317
+ readonly provider: "openai-responses";
307
318
  }, {
308
319
  readonly type: "text";
309
320
  readonly modelName: "gpt-5.2";
@@ -321,7 +332,7 @@ export declare const textModels: readonly [{
321
332
  readonly outputsThinking: false;
322
333
  readonly outputsSignatures: false;
323
334
  };
324
- readonly provider: "openai";
335
+ readonly provider: "openai-responses";
325
336
  }, {
326
337
  readonly type: "text";
327
338
  readonly modelName: "gpt-5.2-pro";
@@ -335,7 +346,7 @@ export declare const textModels: readonly [{
335
346
  readonly outputsThinking: false;
336
347
  readonly outputsSignatures: false;
337
348
  };
338
- readonly provider: "openai";
349
+ readonly provider: "openai-responses";
339
350
  }, {
340
351
  readonly type: "text";
341
352
  readonly modelName: "gpt-5.4";
@@ -352,7 +363,7 @@ export declare const textModels: readonly [{
352
363
  readonly outputsThinking: false;
353
364
  readonly outputsSignatures: false;
354
365
  };
355
- readonly provider: "openai";
366
+ readonly provider: "openai-responses";
356
367
  }, {
357
368
  readonly type: "text";
358
369
  readonly modelName: "gpt-5.4-mini";
@@ -370,7 +381,7 @@ export declare const textModels: readonly [{
370
381
  readonly outputsThinking: false;
371
382
  readonly outputsSignatures: false;
372
383
  };
373
- readonly provider: "openai";
384
+ readonly provider: "openai-responses";
374
385
  }, {
375
386
  readonly type: "text";
376
387
  readonly modelName: "gpt-5.4-nano";
@@ -388,7 +399,7 @@ export declare const textModels: readonly [{
388
399
  readonly outputsThinking: false;
389
400
  readonly outputsSignatures: false;
390
401
  };
391
- readonly provider: "openai";
402
+ readonly provider: "openai-responses";
392
403
  }, {
393
404
  readonly type: "text";
394
405
  readonly modelName: "gpt-5.4-pro";
@@ -404,7 +415,7 @@ export declare const textModels: readonly [{
404
415
  readonly outputsThinking: false;
405
416
  readonly outputsSignatures: false;
406
417
  };
407
- readonly provider: "openai";
418
+ readonly provider: "openai-responses";
408
419
  }, {
409
420
  readonly type: "text";
410
421
  readonly modelName: "gpt-5.5";
@@ -422,7 +433,7 @@ export declare const textModels: readonly [{
422
433
  readonly outputsThinking: false;
423
434
  readonly outputsSignatures: false;
424
435
  };
425
- readonly provider: "openai";
436
+ readonly provider: "openai-responses";
426
437
  }, {
427
438
  readonly type: "text";
428
439
  readonly modelName: "gpt-5.5-pro";
@@ -438,7 +449,7 @@ export declare const textModels: readonly [{
438
449
  readonly outputsThinking: false;
439
450
  readonly outputsSignatures: false;
440
451
  };
441
- readonly provider: "openai";
452
+ readonly provider: "openai-responses";
442
453
  }, {
443
454
  readonly type: "text";
444
455
  readonly modelName: "gemini-3.1-pro-preview";
@@ -466,10 +477,27 @@ export declare const textModels: readonly [{
466
477
  readonly outputTokenCost: 12;
467
478
  readonly disabled: true;
468
479
  readonly provider: "google";
480
+ }, {
481
+ readonly type: "text";
482
+ readonly modelName: "gemini-3.5-flash";
483
+ readonly description: "Latest Gemini 3.5 Flash model (GA May 2026). Outperforms Gemini 3.1 Pro on coding and agentic suites at 4x the speed. 1M context window, 64K output. Context caching: $0.15/1M read.";
484
+ readonly maxInputTokens: 1048576;
485
+ readonly maxOutputTokens: 65536;
486
+ readonly inputTokenCost: 1.5;
487
+ readonly cachedInputTokenCost: 0.15;
488
+ readonly outputTokenCost: 9;
489
+ readonly reasoning: {
490
+ readonly levels: readonly ["minimal", "low", "medium", "high"];
491
+ readonly defaultLevel: "high";
492
+ readonly canDisable: false;
493
+ readonly outputsThinking: true;
494
+ readonly outputsSignatures: true;
495
+ };
496
+ readonly provider: "google";
469
497
  }, {
470
498
  readonly type: "text";
471
499
  readonly modelName: "gemini-3-flash-preview";
472
- readonly description: "Latest Gemini 3 flash model with 1M context window and 64K output. Outperforms 2.5 Pro while being 3x faster. Optimized for agentic workflows and coding. Includes context caching for 90% cost reductions.";
500
+ readonly description: "Gemini 3 Flash preview. Superseded by gemini-3.5-flash. 1M context window and 64K output. Optimized for agentic workflows and coding.";
473
501
  readonly maxInputTokens: 1048576;
474
502
  readonly maxOutputTokens: 65536;
475
503
  readonly inputTokenCost: 0.5;
@@ -485,8 +513,8 @@ export declare const textModels: readonly [{
485
513
  readonly provider: "google";
486
514
  }, {
487
515
  readonly type: "text";
488
- readonly modelName: "gemini-3.1-flash-lite-preview";
489
- readonly description: "Most cost-effective Gemini 3.1 model with thinking support and 1M context window. 2.5x faster TTFA and 45% faster output than 2.5 Flash. Released March 2026.";
516
+ readonly modelName: "gemini-3.1-flash-lite";
517
+ readonly description: "Most cost-effective Gemini 3.1 model (GA). Thinking support, 1M context window, 64K output. 2.5x faster TTFA and 45% faster output than 2.5 Flash.";
490
518
  readonly maxInputTokens: 1048576;
491
519
  readonly maxOutputTokens: 65536;
492
520
  readonly inputTokenCost: 0.25;
@@ -500,6 +528,16 @@ export declare const textModels: readonly [{
500
528
  readonly outputsSignatures: true;
501
529
  };
502
530
  readonly provider: "google";
531
+ }, {
532
+ readonly type: "text";
533
+ readonly modelName: "gemini-3.1-flash-lite-preview";
534
+ readonly description: "DEPRECATED: Preview version, discontinued July 9, 2026. Use gemini-3.1-flash-lite instead.";
535
+ readonly maxInputTokens: 1048576;
536
+ readonly maxOutputTokens: 65536;
537
+ readonly inputTokenCost: 0.25;
538
+ readonly outputTokenCost: 1.5;
539
+ readonly disabled: true;
540
+ readonly provider: "google";
503
541
  }, {
504
542
  readonly type: "text";
505
543
  readonly modelName: "gemini-2.5-pro";
@@ -507,6 +545,7 @@ export declare const textModels: readonly [{
507
545
  readonly maxInputTokens: 2097152;
508
546
  readonly maxOutputTokens: 65536;
509
547
  readonly inputTokenCost: 1.25;
548
+ readonly cachedInputTokenCost: 0.31;
510
549
  readonly outputTokenCost: 10;
511
550
  readonly outputTokensPerSecond: 145;
512
551
  readonly reasoning: {
@@ -522,6 +561,7 @@ export declare const textModels: readonly [{
522
561
  readonly maxInputTokens: 1048576;
523
562
  readonly maxOutputTokens: 65536;
524
563
  readonly inputTokenCost: 0.3;
564
+ readonly cachedInputTokenCost: 0.075;
525
565
  readonly outputTokenCost: 2.5;
526
566
  readonly outputTokensPerSecond: 245;
527
567
  readonly reasoning: {
@@ -537,6 +577,7 @@ export declare const textModels: readonly [{
537
577
  readonly maxInputTokens: 1048576;
538
578
  readonly maxOutputTokens: 65536;
539
579
  readonly inputTokenCost: 0.1;
580
+ readonly cachedInputTokenCost: 0.025;
540
581
  readonly outputTokenCost: 0.4;
541
582
  readonly outputTokensPerSecond: 400;
542
583
  readonly reasoning: {
@@ -611,17 +652,35 @@ export declare const textModels: readonly [{
611
652
  readonly costUnit: "characters";
612
653
  readonly disabled: true;
613
654
  readonly provider: "google";
655
+ }, {
656
+ readonly type: "text";
657
+ readonly modelName: "claude-opus-4-8";
658
+ readonly description: "The most capable Claude model for complex reasoning and agentic coding. Same per-token pricing as Opus 4.7 with improved tool-use efficiency (~290 tokens for tool-use system prompt vs 675 on 4.7). 1M context window, 128K max output.";
659
+ readonly maxInputTokens: 1000000;
660
+ readonly maxOutputTokens: 128000;
661
+ readonly inputTokenCost: 5;
662
+ readonly cachedInputTokenCost: 0.5;
663
+ readonly outputTokenCost: 25;
664
+ readonly reasoning: {
665
+ readonly thinkingStyle: "adaptive";
666
+ readonly canDisable: false;
667
+ readonly outputsThinking: true;
668
+ readonly outputsSignatures: true;
669
+ };
670
+ readonly provider: "anthropic";
614
671
  }, {
615
672
  readonly type: "text";
616
673
  readonly modelName: "claude-opus-4-7";
617
- readonly description: "The most capable Claude model for complex reasoning and agentic coding. Features Adaptive Thinking that auto-tunes reasoning compute per request. 1M context window, 128K max output. Knowledge cutoff: January 2026.";
674
+ readonly description: "Claude Opus 4.7 for complex reasoning and agentic coding. Features Adaptive Thinking that auto-tunes reasoning compute per request. 1M context window, 128K max output. Knowledge cutoff: January 2026.";
618
675
  readonly maxInputTokens: 1000000;
619
676
  readonly maxOutputTokens: 128000;
620
677
  readonly inputTokenCost: 5;
621
678
  readonly cachedInputTokenCost: 0.5;
679
+ readonly cacheCreationInputTokenCost: 6.25;
622
680
  readonly outputTokenCost: 25;
623
681
  readonly outputTokensPerSecond: 72;
624
682
  readonly reasoning: {
683
+ readonly thinkingStyle: "adaptive";
625
684
  readonly canDisable: false;
626
685
  readonly outputsThinking: true;
627
686
  readonly outputsSignatures: true;
@@ -635,9 +694,11 @@ export declare const textModels: readonly [{
635
694
  readonly maxOutputTokens: 128000;
636
695
  readonly inputTokenCost: 5;
637
696
  readonly cachedInputTokenCost: 0.5;
697
+ readonly cacheCreationInputTokenCost: 6.25;
638
698
  readonly outputTokenCost: 25;
639
699
  readonly outputTokensPerSecond: 53;
640
700
  readonly reasoning: {
701
+ readonly thinkingStyle: "adaptive";
641
702
  readonly canDisable: true;
642
703
  readonly outputsThinking: true;
643
704
  readonly outputsSignatures: true;
@@ -651,9 +712,11 @@ export declare const textModels: readonly [{
651
712
  readonly maxOutputTokens: 64000;
652
713
  readonly inputTokenCost: 3;
653
714
  readonly cachedInputTokenCost: 0.3;
715
+ readonly cacheCreationInputTokenCost: 3.75;
654
716
  readonly outputTokenCost: 15;
655
717
  readonly outputTokensPerSecond: 52;
656
718
  readonly reasoning: {
719
+ readonly thinkingStyle: "adaptive";
657
720
  readonly canDisable: true;
658
721
  readonly outputsThinking: true;
659
722
  readonly outputsSignatures: true;
@@ -667,9 +730,11 @@ export declare const textModels: readonly [{
667
730
  readonly maxOutputTokens: 64000;
668
731
  readonly inputTokenCost: 1;
669
732
  readonly cachedInputTokenCost: 0.1;
733
+ readonly cacheCreationInputTokenCost: 1.25;
670
734
  readonly outputTokenCost: 5;
671
735
  readonly outputTokensPerSecond: 97;
672
736
  readonly reasoning: {
737
+ readonly thinkingStyle: "budget";
673
738
  readonly canDisable: true;
674
739
  readonly outputsThinking: true;
675
740
  readonly outputsSignatures: true;
@@ -685,6 +750,7 @@ export declare const textModels: readonly [{
685
750
  readonly outputTokenCost: 15;
686
751
  readonly outputTokensPerSecond: 78;
687
752
  readonly reasoning: {
753
+ readonly thinkingStyle: "budget";
688
754
  readonly canDisable: true;
689
755
  readonly outputsThinking: true;
690
756
  readonly outputsSignatures: true;
@@ -734,7 +800,14 @@ export declare const imageModels: readonly [{
734
800
  readonly type: "image";
735
801
  readonly modelName: "gemini-3.1-flash-image-preview";
736
802
  readonly provider: "google";
737
- readonly description: "Fast image generation with Gemini 3.1 Flash. Supports resolutions from 512px to 4096px. ~$0.067/image at 1K resolution.";
803
+ readonly description: "DEPRECATED: Preview version. Use gemini-3.1-flash-image instead.";
804
+ readonly costPerImage: 0.067;
805
+ readonly disabled: true;
806
+ }, {
807
+ readonly type: "image";
808
+ readonly modelName: "gemini-3.1-flash-image";
809
+ readonly provider: "google";
810
+ readonly description: "Fast image generation with Gemini 3.1 Flash (GA). Supports resolutions from 512px to 4096px. ~$0.045/image at 512px, $0.067 at 1K, $0.101 at 2K, $0.151 at 4K.";
738
811
  readonly costPerImage: 0.067;
739
812
  }];
740
813
  export declare const embeddingsModels: EmbeddingsModel[];
@@ -940,7 +1013,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
940
1013
  readonly outputsThinking: false;
941
1014
  readonly outputsSignatures: false;
942
1015
  };
943
- readonly provider: "openai";
1016
+ readonly provider: "openai-responses";
944
1017
  } | {
945
1018
  readonly type: "text";
946
1019
  readonly modelName: "gpt-5-mini";
@@ -958,7 +1031,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
958
1031
  readonly outputsThinking: false;
959
1032
  readonly outputsSignatures: false;
960
1033
  };
961
- readonly provider: "openai";
1034
+ readonly provider: "openai-responses";
962
1035
  } | {
963
1036
  readonly type: "text";
964
1037
  readonly modelName: "gpt-5-nano";
@@ -976,7 +1049,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
976
1049
  readonly outputsThinking: false;
977
1050
  readonly outputsSignatures: false;
978
1051
  };
979
- readonly provider: "openai";
1052
+ readonly provider: "openai-responses";
980
1053
  } | {
981
1054
  readonly type: "text";
982
1055
  readonly modelName: "gpt-5.1";
@@ -993,7 +1066,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
993
1066
  readonly outputsThinking: false;
994
1067
  readonly outputsSignatures: false;
995
1068
  };
996
- readonly provider: "openai";
1069
+ readonly provider: "openai-responses";
997
1070
  } | {
998
1071
  readonly type: "text";
999
1072
  readonly modelName: "gpt-5.2";
@@ -1011,7 +1084,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1011
1084
  readonly outputsThinking: false;
1012
1085
  readonly outputsSignatures: false;
1013
1086
  };
1014
- readonly provider: "openai";
1087
+ readonly provider: "openai-responses";
1015
1088
  } | {
1016
1089
  readonly type: "text";
1017
1090
  readonly modelName: "gpt-5.2-pro";
@@ -1025,7 +1098,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1025
1098
  readonly outputsThinking: false;
1026
1099
  readonly outputsSignatures: false;
1027
1100
  };
1028
- readonly provider: "openai";
1101
+ readonly provider: "openai-responses";
1029
1102
  } | {
1030
1103
  readonly type: "text";
1031
1104
  readonly modelName: "gpt-5.4";
@@ -1042,7 +1115,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1042
1115
  readonly outputsThinking: false;
1043
1116
  readonly outputsSignatures: false;
1044
1117
  };
1045
- readonly provider: "openai";
1118
+ readonly provider: "openai-responses";
1046
1119
  } | {
1047
1120
  readonly type: "text";
1048
1121
  readonly modelName: "gpt-5.4-mini";
@@ -1060,7 +1133,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1060
1133
  readonly outputsThinking: false;
1061
1134
  readonly outputsSignatures: false;
1062
1135
  };
1063
- readonly provider: "openai";
1136
+ readonly provider: "openai-responses";
1064
1137
  } | {
1065
1138
  readonly type: "text";
1066
1139
  readonly modelName: "gpt-5.4-nano";
@@ -1078,7 +1151,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1078
1151
  readonly outputsThinking: false;
1079
1152
  readonly outputsSignatures: false;
1080
1153
  };
1081
- readonly provider: "openai";
1154
+ readonly provider: "openai-responses";
1082
1155
  } | {
1083
1156
  readonly type: "text";
1084
1157
  readonly modelName: "gpt-5.4-pro";
@@ -1094,7 +1167,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1094
1167
  readonly outputsThinking: false;
1095
1168
  readonly outputsSignatures: false;
1096
1169
  };
1097
- readonly provider: "openai";
1170
+ readonly provider: "openai-responses";
1098
1171
  } | {
1099
1172
  readonly type: "text";
1100
1173
  readonly modelName: "gpt-5.5";
@@ -1112,7 +1185,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1112
1185
  readonly outputsThinking: false;
1113
1186
  readonly outputsSignatures: false;
1114
1187
  };
1115
- readonly provider: "openai";
1188
+ readonly provider: "openai-responses";
1116
1189
  } | {
1117
1190
  readonly type: "text";
1118
1191
  readonly modelName: "gpt-5.5-pro";
@@ -1128,7 +1201,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1128
1201
  readonly outputsThinking: false;
1129
1202
  readonly outputsSignatures: false;
1130
1203
  };
1131
- readonly provider: "openai";
1204
+ readonly provider: "openai-responses";
1132
1205
  } | {
1133
1206
  readonly type: "text";
1134
1207
  readonly modelName: "gemini-3.1-pro-preview";
@@ -1156,10 +1229,27 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1156
1229
  readonly outputTokenCost: 12;
1157
1230
  readonly disabled: true;
1158
1231
  readonly provider: "google";
1232
+ } | {
1233
+ readonly type: "text";
1234
+ readonly modelName: "gemini-3.5-flash";
1235
+ readonly description: "Latest Gemini 3.5 Flash model (GA May 2026). Outperforms Gemini 3.1 Pro on coding and agentic suites at 4x the speed. 1M context window, 64K output. Context caching: $0.15/1M read.";
1236
+ readonly maxInputTokens: 1048576;
1237
+ readonly maxOutputTokens: 65536;
1238
+ readonly inputTokenCost: 1.5;
1239
+ readonly cachedInputTokenCost: 0.15;
1240
+ readonly outputTokenCost: 9;
1241
+ readonly reasoning: {
1242
+ readonly levels: readonly ["minimal", "low", "medium", "high"];
1243
+ readonly defaultLevel: "high";
1244
+ readonly canDisable: false;
1245
+ readonly outputsThinking: true;
1246
+ readonly outputsSignatures: true;
1247
+ };
1248
+ readonly provider: "google";
1159
1249
  } | {
1160
1250
  readonly type: "text";
1161
1251
  readonly modelName: "gemini-3-flash-preview";
1162
- readonly description: "Latest Gemini 3 flash model with 1M context window and 64K output. Outperforms 2.5 Pro while being 3x faster. Optimized for agentic workflows and coding. Includes context caching for 90% cost reductions.";
1252
+ readonly description: "Gemini 3 Flash preview. Superseded by gemini-3.5-flash. 1M context window and 64K output. Optimized for agentic workflows and coding.";
1163
1253
  readonly maxInputTokens: 1048576;
1164
1254
  readonly maxOutputTokens: 65536;
1165
1255
  readonly inputTokenCost: 0.5;
@@ -1175,8 +1265,8 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1175
1265
  readonly provider: "google";
1176
1266
  } | {
1177
1267
  readonly type: "text";
1178
- readonly modelName: "gemini-3.1-flash-lite-preview";
1179
- readonly description: "Most cost-effective Gemini 3.1 model with thinking support and 1M context window. 2.5x faster TTFA and 45% faster output than 2.5 Flash. Released March 2026.";
1268
+ readonly modelName: "gemini-3.1-flash-lite";
1269
+ readonly description: "Most cost-effective Gemini 3.1 model (GA). Thinking support, 1M context window, 64K output. 2.5x faster TTFA and 45% faster output than 2.5 Flash.";
1180
1270
  readonly maxInputTokens: 1048576;
1181
1271
  readonly maxOutputTokens: 65536;
1182
1272
  readonly inputTokenCost: 0.25;
@@ -1190,6 +1280,16 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1190
1280
  readonly outputsSignatures: true;
1191
1281
  };
1192
1282
  readonly provider: "google";
1283
+ } | {
1284
+ readonly type: "text";
1285
+ readonly modelName: "gemini-3.1-flash-lite-preview";
1286
+ readonly description: "DEPRECATED: Preview version, discontinued July 9, 2026. Use gemini-3.1-flash-lite instead.";
1287
+ readonly maxInputTokens: 1048576;
1288
+ readonly maxOutputTokens: 65536;
1289
+ readonly inputTokenCost: 0.25;
1290
+ readonly outputTokenCost: 1.5;
1291
+ readonly disabled: true;
1292
+ readonly provider: "google";
1193
1293
  } | {
1194
1294
  readonly type: "text";
1195
1295
  readonly modelName: "gemini-2.5-pro";
@@ -1197,6 +1297,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1197
1297
  readonly maxInputTokens: 2097152;
1198
1298
  readonly maxOutputTokens: 65536;
1199
1299
  readonly inputTokenCost: 1.25;
1300
+ readonly cachedInputTokenCost: 0.31;
1200
1301
  readonly outputTokenCost: 10;
1201
1302
  readonly outputTokensPerSecond: 145;
1202
1303
  readonly reasoning: {
@@ -1212,6 +1313,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1212
1313
  readonly maxInputTokens: 1048576;
1213
1314
  readonly maxOutputTokens: 65536;
1214
1315
  readonly inputTokenCost: 0.3;
1316
+ readonly cachedInputTokenCost: 0.075;
1215
1317
  readonly outputTokenCost: 2.5;
1216
1318
  readonly outputTokensPerSecond: 245;
1217
1319
  readonly reasoning: {
@@ -1227,6 +1329,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1227
1329
  readonly maxInputTokens: 1048576;
1228
1330
  readonly maxOutputTokens: 65536;
1229
1331
  readonly inputTokenCost: 0.1;
1332
+ readonly cachedInputTokenCost: 0.025;
1230
1333
  readonly outputTokenCost: 0.4;
1231
1334
  readonly outputTokensPerSecond: 400;
1232
1335
  readonly reasoning: {
@@ -1301,17 +1404,35 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1301
1404
  readonly costUnit: "characters";
1302
1405
  readonly disabled: true;
1303
1406
  readonly provider: "google";
1407
+ } | {
1408
+ readonly type: "text";
1409
+ readonly modelName: "claude-opus-4-8";
1410
+ readonly description: "The most capable Claude model for complex reasoning and agentic coding. Same per-token pricing as Opus 4.7 with improved tool-use efficiency (~290 tokens for tool-use system prompt vs 675 on 4.7). 1M context window, 128K max output.";
1411
+ readonly maxInputTokens: 1000000;
1412
+ readonly maxOutputTokens: 128000;
1413
+ readonly inputTokenCost: 5;
1414
+ readonly cachedInputTokenCost: 0.5;
1415
+ readonly outputTokenCost: 25;
1416
+ readonly reasoning: {
1417
+ readonly thinkingStyle: "adaptive";
1418
+ readonly canDisable: false;
1419
+ readonly outputsThinking: true;
1420
+ readonly outputsSignatures: true;
1421
+ };
1422
+ readonly provider: "anthropic";
1304
1423
  } | {
1305
1424
  readonly type: "text";
1306
1425
  readonly modelName: "claude-opus-4-7";
1307
- readonly description: "The most capable Claude model for complex reasoning and agentic coding. Features Adaptive Thinking that auto-tunes reasoning compute per request. 1M context window, 128K max output. Knowledge cutoff: January 2026.";
1426
+ readonly description: "Claude Opus 4.7 for complex reasoning and agentic coding. Features Adaptive Thinking that auto-tunes reasoning compute per request. 1M context window, 128K max output. Knowledge cutoff: January 2026.";
1308
1427
  readonly maxInputTokens: 1000000;
1309
1428
  readonly maxOutputTokens: 128000;
1310
1429
  readonly inputTokenCost: 5;
1311
1430
  readonly cachedInputTokenCost: 0.5;
1431
+ readonly cacheCreationInputTokenCost: 6.25;
1312
1432
  readonly outputTokenCost: 25;
1313
1433
  readonly outputTokensPerSecond: 72;
1314
1434
  readonly reasoning: {
1435
+ readonly thinkingStyle: "adaptive";
1315
1436
  readonly canDisable: false;
1316
1437
  readonly outputsThinking: true;
1317
1438
  readonly outputsSignatures: true;
@@ -1325,9 +1446,11 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1325
1446
  readonly maxOutputTokens: 128000;
1326
1447
  readonly inputTokenCost: 5;
1327
1448
  readonly cachedInputTokenCost: 0.5;
1449
+ readonly cacheCreationInputTokenCost: 6.25;
1328
1450
  readonly outputTokenCost: 25;
1329
1451
  readonly outputTokensPerSecond: 53;
1330
1452
  readonly reasoning: {
1453
+ readonly thinkingStyle: "adaptive";
1331
1454
  readonly canDisable: true;
1332
1455
  readonly outputsThinking: true;
1333
1456
  readonly outputsSignatures: true;
@@ -1341,9 +1464,11 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1341
1464
  readonly maxOutputTokens: 64000;
1342
1465
  readonly inputTokenCost: 3;
1343
1466
  readonly cachedInputTokenCost: 0.3;
1467
+ readonly cacheCreationInputTokenCost: 3.75;
1344
1468
  readonly outputTokenCost: 15;
1345
1469
  readonly outputTokensPerSecond: 52;
1346
1470
  readonly reasoning: {
1471
+ readonly thinkingStyle: "adaptive";
1347
1472
  readonly canDisable: true;
1348
1473
  readonly outputsThinking: true;
1349
1474
  readonly outputsSignatures: true;
@@ -1357,9 +1482,11 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1357
1482
  readonly maxOutputTokens: 64000;
1358
1483
  readonly inputTokenCost: 1;
1359
1484
  readonly cachedInputTokenCost: 0.1;
1485
+ readonly cacheCreationInputTokenCost: 1.25;
1360
1486
  readonly outputTokenCost: 5;
1361
1487
  readonly outputTokensPerSecond: 97;
1362
1488
  readonly reasoning: {
1489
+ readonly thinkingStyle: "budget";
1363
1490
  readonly canDisable: true;
1364
1491
  readonly outputsThinking: true;
1365
1492
  readonly outputsSignatures: true;
@@ -1375,6 +1502,7 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1375
1502
  readonly outputTokenCost: 15;
1376
1503
  readonly outputTokensPerSecond: 78;
1377
1504
  readonly reasoning: {
1505
+ readonly thinkingStyle: "budget";
1378
1506
  readonly canDisable: true;
1379
1507
  readonly outputsThinking: true;
1380
1508
  readonly outputsSignatures: true;
@@ -1423,7 +1551,14 @@ export declare function getModel(modelName: ModelName): TextModel | EmbeddingsMo
1423
1551
  readonly type: "image";
1424
1552
  readonly modelName: "gemini-3.1-flash-image-preview";
1425
1553
  readonly provider: "google";
1426
- readonly description: "Fast image generation with Gemini 3.1 Flash. Supports resolutions from 512px to 4096px. ~$0.067/image at 1K resolution.";
1554
+ readonly description: "DEPRECATED: Preview version. Use gemini-3.1-flash-image instead.";
1555
+ readonly costPerImage: 0.067;
1556
+ readonly disabled: true;
1557
+ } | {
1558
+ readonly type: "image";
1559
+ readonly modelName: "gemini-3.1-flash-image";
1560
+ readonly provider: "google";
1561
+ readonly description: "Fast image generation with Gemini 3.1 Flash (GA). Supports resolutions from 512px to 4096px. ~$0.045/image at 512px, $0.067 at 1K, $0.101 at 2K, $0.151 at 4K.";
1427
1562
  readonly costPerImage: 0.067;
1428
1563
  } | undefined;
1429
1564
  export declare function isImageModel(model: ModelType): model is ImageModel;