@just-every/ensemble 0.2.161 → 0.2.162

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.
@@ -8,8 +8,12 @@ exports.MODEL_CLASSES = {
8
8
  models: [
9
9
  'gpt-5-mini',
10
10
  'gemini-2.5-flash-preview-05-20',
11
- 'claude-3-5-haiku-latest',
11
+ 'gemini-2.5-flash',
12
+ 'gemini-3-pro',
13
+ 'claude-sonnet-4-5-20250514',
12
14
  'grok-3-mini',
15
+ 'grok-4-1-fast-non-reasoning',
16
+ 'grok-4-fast-non-reasoning',
13
17
  ],
14
18
  random: true,
15
19
  },
@@ -17,7 +21,8 @@ exports.MODEL_CLASSES = {
17
21
  models: [
18
22
  'gpt-5-nano',
19
23
  'gemini-2.5-flash-lite-preview-06-17',
20
- 'claude-3-5-haiku-latest',
24
+ 'gemini-2.5-flash-lite',
25
+ 'claude-haiku-4-5-20250514',
21
26
  'grok-3-mini',
22
27
  'mistral/ministral-8b',
23
28
  'openai/gpt-oss-20b',
@@ -27,9 +32,14 @@ exports.MODEL_CLASSES = {
27
32
  reasoning: {
28
33
  models: [
29
34
  'gpt-5-max',
35
+ 'gpt-5.1',
30
36
  'gemini-2.5-pro-preview-06-05',
31
- 'claude-opus-4-1-20250805-max',
37
+ 'gemini-2.5-pro',
38
+ 'gemini-3-pro',
39
+ 'claude-opus-4-1-20250805',
32
40
  'grok-4',
41
+ 'grok-4-1-fast-reasoning',
42
+ 'grok-4-fast-reasoning',
33
43
  'mistralai/magistral-medium-2506:thinking',
34
44
  'qwen3-235b-a22b-thinking-2507',
35
45
  ],
@@ -38,6 +48,7 @@ exports.MODEL_CLASSES = {
38
48
  reasoning_mini: {
39
49
  models: [
40
50
  'gemini-2.5-flash-preview-05-20-medium',
51
+ 'gemini-2.5-flash',
41
52
  'grok-3-mini',
42
53
  'o3-low',
43
54
  'openai/gpt-oss-120b',
@@ -50,7 +61,7 @@ exports.MODEL_CLASSES = {
50
61
  'gemini-2.5-pro-preview-06-05',
51
62
  'gemini-2.5-flash-preview-05-20-medium',
52
63
  'o3-medium',
53
- 'claude-sonnet-4-20250514-medium',
64
+ 'claude-sonnet-4-5-20250514',
54
65
  ],
55
66
  random: true,
56
67
  },
@@ -60,7 +71,7 @@ exports.MODEL_CLASSES = {
60
71
  'gpt-5-mini',
61
72
  'gemini-2.5-pro-preview-06-05',
62
73
  'gemini-2.5-flash-preview-05-20-high',
63
- 'claude-sonnet-4-20250514-medium',
74
+ 'claude-sonnet-4-5-20250514',
64
75
  'grok-3-mini',
65
76
  ],
66
77
  random: true,
@@ -68,12 +79,15 @@ exports.MODEL_CLASSES = {
68
79
  code: {
69
80
  models: [
70
81
  'gpt-5',
82
+ 'gpt-5.1-codex',
83
+ 'gpt-5.1-codex-mini',
71
84
  'o3-high',
72
85
  'gemini-2.5-pro-preview-06-05-medium',
73
86
  'gemini-2.5-flash-preview-05-20-max',
74
- 'claude-opus-4-1-20250805-medium',
75
- 'claude-sonnet-4-20250514-max',
87
+ 'claude-opus-4-1-20250805',
88
+ 'claude-sonnet-4-5-20250514',
76
89
  'grok-4-medium',
90
+ 'grok-4-1-fast-reasoning',
77
91
  'qwen3-coder',
78
92
  ],
79
93
  random: true,
@@ -97,8 +111,9 @@ exports.MODEL_CLASSES = {
97
111
  'gpt-5',
98
112
  'o3-high',
99
113
  'gemini-2.5-pro-preview-06-05',
100
- 'claude-opus-4-1-20250805-low',
114
+ 'claude-opus-4-1-20250805',
101
115
  'grok-4',
116
+ 'grok-2-vision',
102
117
  ],
103
118
  random: true,
104
119
  },
@@ -108,7 +123,7 @@ exports.MODEL_CLASSES = {
108
123
  'o3-low',
109
124
  'gemini-2.5-flash-lite-preview-06-17',
110
125
  'gemini-2.5-flash-preview-05-20',
111
- 'claude-sonnet-4-20250514-low',
126
+ 'claude-sonnet-4-5-20250514',
112
127
  ],
113
128
  random: true,
114
129
  },
@@ -127,9 +142,13 @@ exports.MODEL_CLASSES = {
127
142
  'gpt-4.1-nano',
128
143
  'gpt-4.1-mini',
129
144
  'gemini-2.5-pro-preview-06-05',
145
+ 'gemini-2.5-pro',
146
+ 'gemini-3-pro',
130
147
  'gemini-2.5-flash-preview-05-20-medium',
131
148
  'gemini-2.5-flash-preview-05-20-low',
149
+ 'gemini-2.5-flash',
132
150
  'gemini-2.5-flash-lite-preview-06-17',
151
+ 'gemini-2.5-flash-lite',
133
152
  ],
134
153
  random: true,
135
154
  description: 'Models with very large context windows (near 1M tokens) for processing long documents',
@@ -138,6 +157,8 @@ exports.MODEL_CLASSES = {
138
157
  models: [
139
158
  'gpt-image-1',
140
159
  'gemini-2.5-flash-image-preview',
160
+ 'gemini-2.5-flash-image',
161
+ 'gemini-3-pro-image-preview',
141
162
  'seedream-4',
142
163
  'luma-photon-1',
143
164
  'luma-photon-flash-1',
@@ -173,6 +194,7 @@ exports.MODEL_CLASSES = {
173
194
  'eleven_flash_v2_5',
174
195
  'gemini-2.5-flash-preview-tts',
175
196
  'gemini-2.5-pro-preview-tts',
197
+ 'gemini-2.5-flash-native-audio-preview',
176
198
  ],
177
199
  description: 'Text-to-Speech models for voice generation',
178
200
  },
@@ -444,15 +466,152 @@ exports.MODEL_REGISTRY = [
444
466
  description: 'Ministral 8B is a state-of-the-art language model optimized for on-device and edge computing. Designed for efficiency in knowledge-intensive tasks, commonsense reasoning, and function-calling.',
445
467
  },
446
468
  {
447
- id: 'grok-3',
448
- aliases: ['grok-3-2025-02-11'],
469
+ id: 'grok-4-1-fast-reasoning',
470
+ aliases: ['grok-4.1-fast-reasoning'],
471
+ provider: 'xai',
472
+ cost: {
473
+ input_per_million: {
474
+ threshold_tokens: 128_000,
475
+ price_below_threshold_per_million: 0.2,
476
+ price_above_threshold_per_million: 0.5,
477
+ },
478
+ output_per_million: {
479
+ threshold_tokens: 128_000,
480
+ price_below_threshold_per_million: 0.5,
481
+ price_above_threshold_per_million: 1.0,
482
+ },
483
+ cached_input_per_million: 0.05,
484
+ },
485
+ features: {
486
+ context_length: 2_000_000,
487
+ input_modality: ['text', 'image'],
488
+ output_modality: ['text'],
489
+ tool_use: true,
490
+ streaming: true,
491
+ json_output: true,
492
+ },
493
+ class: 'reasoning',
494
+ score: 88,
495
+ scores: {
496
+ monologue: 90,
497
+ code: 85,
498
+ reasoning: 88,
499
+ },
500
+ description: 'Grok 4.1 Fast with extended reasoning. 2M context, tiered pricing at 128k threshold.',
501
+ },
502
+ {
503
+ id: 'grok-4-1-fast-non-reasoning',
504
+ aliases: ['grok-4.1-fast-non-reasoning'],
505
+ provider: 'xai',
506
+ cost: {
507
+ input_per_million: {
508
+ threshold_tokens: 128_000,
509
+ price_below_threshold_per_million: 0.2,
510
+ price_above_threshold_per_million: 0.5,
511
+ },
512
+ output_per_million: {
513
+ threshold_tokens: 128_000,
514
+ price_below_threshold_per_million: 0.5,
515
+ price_above_threshold_per_million: 1.0,
516
+ },
517
+ cached_input_per_million: 0.05,
518
+ },
519
+ features: {
520
+ context_length: 2_000_000,
521
+ input_modality: ['text', 'image'],
522
+ output_modality: ['text'],
523
+ tool_use: true,
524
+ streaming: true,
525
+ json_output: true,
526
+ },
527
+ class: 'standard',
528
+ score: 85,
529
+ scores: {
530
+ monologue: 87,
531
+ code: 82,
532
+ reasoning: 80,
533
+ },
534
+ description: 'Grok 4.1 Fast without reasoning. 2M context, tiered pricing at 128k threshold.',
535
+ },
536
+ {
537
+ id: 'grok-4',
538
+ aliases: ['grok-4-2025-09-01'],
449
539
  provider: 'xai',
450
540
  cost: {
451
541
  input_per_million: 3.0,
452
542
  output_per_million: 15.0,
543
+ cached_input_per_million: 0.75,
453
544
  },
454
545
  features: {
455
- context_length: 131_072,
546
+ context_length: 256_000,
547
+ input_modality: ['text', 'image'],
548
+ output_modality: ['text'],
549
+ tool_use: true,
550
+ streaming: true,
551
+ json_output: true,
552
+ },
553
+ class: 'reasoning',
554
+ score: 90,
555
+ scores: {
556
+ monologue: 92,
557
+ code: 88,
558
+ reasoning: 85,
559
+ },
560
+ description: 'Grok-4 premium model with 256k context and vision capabilities.',
561
+ },
562
+ {
563
+ id: 'grok-4-fast-reasoning',
564
+ aliases: ['grok-4-fast-reasoning-2025-09-01'],
565
+ provider: 'xai',
566
+ cost: {
567
+ input_per_million: {
568
+ threshold_tokens: 128_000,
569
+ price_below_threshold_per_million: 0.2,
570
+ price_above_threshold_per_million: 0.5,
571
+ },
572
+ output_per_million: {
573
+ threshold_tokens: 128_000,
574
+ price_below_threshold_per_million: 0.5,
575
+ price_above_threshold_per_million: 1.0,
576
+ },
577
+ cached_input_per_million: 0.05,
578
+ },
579
+ features: {
580
+ context_length: 2_000_000,
581
+ input_modality: ['text', 'image'],
582
+ output_modality: ['text'],
583
+ tool_use: true,
584
+ streaming: true,
585
+ json_output: true,
586
+ },
587
+ class: 'reasoning',
588
+ score: 87,
589
+ scores: {
590
+ monologue: 89,
591
+ code: 84,
592
+ reasoning: 86,
593
+ },
594
+ description: 'Grok-4 Fast with reasoning (Sep 2025). 2M context, tiered pricing at 128k threshold.',
595
+ },
596
+ {
597
+ id: 'grok-4-fast-non-reasoning',
598
+ aliases: ['grok-4-fast-non-reasoning-2025-09-01'],
599
+ provider: 'xai',
600
+ cost: {
601
+ input_per_million: {
602
+ threshold_tokens: 128_000,
603
+ price_below_threshold_per_million: 0.2,
604
+ price_above_threshold_per_million: 0.5,
605
+ },
606
+ output_per_million: {
607
+ threshold_tokens: 128_000,
608
+ price_below_threshold_per_million: 0.5,
609
+ price_above_threshold_per_million: 1.0,
610
+ },
611
+ cached_input_per_million: 0.05,
612
+ },
613
+ features: {
614
+ context_length: 2_000_000,
456
615
  input_modality: ['text', 'image'],
457
616
  output_modality: ['text'],
458
617
  tool_use: true,
@@ -460,21 +619,21 @@ exports.MODEL_REGISTRY = [
460
619
  json_output: true,
461
620
  },
462
621
  class: 'standard',
463
- score: 78,
622
+ score: 84,
464
623
  scores: {
465
- monologue: 80,
466
- code: 70,
467
- reasoning: 65,
624
+ monologue: 86,
625
+ code: 81,
626
+ reasoning: 79,
468
627
  },
469
- description: 'Flagship Grok-3 model for complex reasoning and generation',
628
+ description: 'Grok-4 Fast without reasoning (Sep 2025). 2M context, tiered pricing at 128k threshold.',
470
629
  },
471
630
  {
472
631
  id: 'grok-3',
473
- aliases: ['grok-3-2025-04-11'],
632
+ aliases: ['grok-3-2025-02-11'],
474
633
  provider: 'xai',
475
634
  cost: {
476
- input_per_million: 5.0,
477
- output_per_million: 25.0,
635
+ input_per_million: 3.0,
636
+ output_per_million: 15.0,
478
637
  },
479
638
  features: {
480
639
  context_length: 131_072,
@@ -491,7 +650,7 @@ exports.MODEL_REGISTRY = [
491
650
  code: 70,
492
651
  reasoning: 65,
493
652
  },
494
- description: 'Same Grok-3 weights on premium infra for lower latency',
653
+ description: 'Grok-3 model with 131k context.',
495
654
  },
496
655
  {
497
656
  id: 'grok-3-mini',
@@ -516,11 +675,11 @@ exports.MODEL_REGISTRY = [
516
675
  code: 55,
517
676
  reasoning: 50,
518
677
  },
519
- description: 'Lightweight Grok-3 Mini—budget model for logic tasks',
678
+ description: 'Grok-3 Mini—budget model with 131k context.',
520
679
  },
521
680
  {
522
- id: 'grok-3-mini',
523
- aliases: ['grok-3-mini-2025-04-11'],
681
+ id: 'grok-3-mini-accelerated',
682
+ aliases: ['grok-3-mini-accelerated-2025-04-11'],
524
683
  provider: 'xai',
525
684
  cost: {
526
685
  input_per_million: 0.6,
@@ -541,30 +700,57 @@ exports.MODEL_REGISTRY = [
541
700
  code: 55,
542
701
  reasoning: 50,
543
702
  },
544
- description: 'Grok-3 Mini on accelerated hardware for latency-critical use',
703
+ description: 'Grok-3 Mini on accelerated hardware for lower latency.',
545
704
  },
546
705
  {
547
- id: 'grok-4',
548
- aliases: ['grok-4-2025-07-09'],
706
+ id: 'grok-2',
707
+ aliases: ['grok-2-text'],
549
708
  provider: 'xai',
550
709
  cost: {
551
- input_per_million: 10.0,
552
- output_per_million: 30.0,
710
+ input_per_million: 2.0,
711
+ output_per_million: 10.0,
553
712
  },
554
713
  features: {
555
- context_length: 131072,
714
+ context_length: 128_000,
715
+ input_modality: ['text'],
716
+ output_modality: ['text'],
717
+ tool_use: true,
718
+ streaming: true,
719
+ json_output: true,
720
+ },
721
+ class: 'standard',
722
+ score: 70,
723
+ scores: {
724
+ monologue: 72,
725
+ code: 65,
726
+ reasoning: 60,
727
+ },
728
+ description: 'Legacy Grok-2 text-only model.',
729
+ },
730
+ {
731
+ id: 'grok-2-vision',
732
+ aliases: ['grok-2-vision'],
733
+ provider: 'xai',
734
+ cost: {
735
+ input_per_million: 2.0,
736
+ output_per_million: 10.0,
737
+ },
738
+ features: {
739
+ context_length: 128_000,
556
740
  input_modality: ['text', 'image'],
557
741
  output_modality: ['text'],
558
742
  tool_use: true,
559
743
  streaming: true,
560
744
  json_output: true,
561
745
  },
746
+ class: 'vision',
747
+ score: 70,
562
748
  scores: {
563
- monologue: 85,
564
- code: 75,
565
- reasoning: 70,
749
+ monologue: 72,
750
+ code: 65,
751
+ reasoning: 60,
566
752
  },
567
- description: 'Most advanced Grok model with vision capabilities and complex reasoning',
753
+ description: 'Legacy Grok-2 model with vision capabilities.',
568
754
  },
569
755
  {
570
756
  id: 'gpt-4.1',
@@ -758,6 +944,89 @@ exports.MODEL_REGISTRY = [
758
944
  },
759
945
  description: 'GPT-5 model used in ChatGPT',
760
946
  },
947
+ {
948
+ id: 'gpt-5.1',
949
+ aliases: ['gpt-5.1-chat-latest'],
950
+ provider: 'openai',
951
+ cost: {
952
+ input_per_million: 1.25,
953
+ output_per_million: 10.0,
954
+ cached_input_per_million: 0.125,
955
+ },
956
+ features: {
957
+ context_length: 256000,
958
+ input_modality: ['text', 'image', 'audio', 'video'],
959
+ output_modality: ['text', 'image', 'audio'],
960
+ tool_use: true,
961
+ streaming: true,
962
+ json_output: true,
963
+ },
964
+ class: 'reasoning',
965
+ score: 96,
966
+ description: 'GPT-5.1 flagship multimodal model (Instant/Thinking modes, pricing TBD).',
967
+ },
968
+ {
969
+ id: 'gpt-5.1-codex',
970
+ provider: 'openai',
971
+ cost: {
972
+ input_per_million: 1.25,
973
+ output_per_million: 10.0,
974
+ cached_input_per_million: 0.125,
975
+ },
976
+ features: {
977
+ context_length: 400000,
978
+ input_modality: ['text'],
979
+ output_modality: ['text'],
980
+ tool_use: true,
981
+ streaming: true,
982
+ json_output: true,
983
+ max_output_tokens: 128000,
984
+ },
985
+ class: 'code',
986
+ score: 92,
987
+ description: 'GPT-5.1 Codex — optimized for agentic coding tasks (400k/128k).',
988
+ },
989
+ {
990
+ id: 'gpt-5.1-codex-mini',
991
+ provider: 'openai',
992
+ cost: {
993
+ input_per_million: 0.25,
994
+ output_per_million: 2.0,
995
+ cached_input_per_million: 0.025,
996
+ },
997
+ features: {
998
+ context_length: 400000,
999
+ input_modality: ['text'],
1000
+ output_modality: ['text'],
1001
+ tool_use: true,
1002
+ streaming: true,
1003
+ json_output: true,
1004
+ max_output_tokens: 128000,
1005
+ },
1006
+ class: 'code',
1007
+ score: 86,
1008
+ description: 'GPT-5.1 Codex Mini — lightweight coding model with 400k context.',
1009
+ },
1010
+ {
1011
+ id: 'gpt-5.1-codex-max',
1012
+ provider: 'openai',
1013
+ cost: {
1014
+ input_per_million: 5.0,
1015
+ output_per_million: 15.0,
1016
+ },
1017
+ features: {
1018
+ context_length: 10_000_000,
1019
+ input_modality: ['text'],
1020
+ output_modality: ['text'],
1021
+ tool_use: true,
1022
+ streaming: true,
1023
+ json_output: true,
1024
+ max_output_tokens: 128000,
1025
+ },
1026
+ class: 'code',
1027
+ score: 95,
1028
+ description: 'GPT-5.1 Codex Max — multi-million context (IDE-only for now), pricing TBD.',
1029
+ },
761
1030
  {
762
1031
  id: 'gpt-4o',
763
1032
  aliases: ['gpt-4o-2024-08-06'],
@@ -987,8 +1256,8 @@ exports.MODEL_REGISTRY = [
987
1256
  description: 'Model that can understand and control computer interfaces',
988
1257
  },
989
1258
  {
990
- id: 'claude-3-7-sonnet-latest',
991
- aliases: ['claude-3-7-sonnet'],
1259
+ id: 'claude-sonnet-4-5-20250514',
1260
+ aliases: ['claude-sonnet-4-5', 'claude-sonnet-4.5', 'claude-sonnet-latest'],
992
1261
  provider: 'anthropic',
993
1262
  cost: {
994
1263
  input_per_million: 3.0,
@@ -1002,25 +1271,84 @@ exports.MODEL_REGISTRY = [
1002
1271
  tool_use: true,
1003
1272
  streaming: true,
1004
1273
  json_output: true,
1274
+ reasoning_output: true,
1005
1275
  max_output_tokens: 64000,
1006
1276
  },
1007
1277
  class: 'reasoning',
1008
- score: 85,
1278
+ score: 88,
1009
1279
  scores: {
1010
- monologue: 83,
1011
- code: 77,
1012
- reasoning: 69,
1280
+ monologue: 86,
1281
+ code: 85,
1282
+ reasoning: 84,
1013
1283
  },
1014
- description: 'Latest Claude model with strong reasoning capabilities (extended thinking internal)',
1284
+ description: 'Claude Sonnet 4.5 - Latest general-purpose model with strong reasoning and text/image support',
1015
1285
  },
1016
1286
  {
1017
- id: 'claude-3-5-haiku-latest',
1018
- aliases: ['claude-3-5-haiku'],
1287
+ id: 'claude-sonnet-4-5-20250929',
1288
+ aliases: ['claude-sonnet-4-5-sep-2025', 'claude-sonnet-4.5-2025-09-29'],
1019
1289
  provider: 'anthropic',
1020
1290
  cost: {
1021
- input_per_million: 0.8,
1022
- output_per_million: 4.0,
1023
- cached_input_per_million: 0.08,
1291
+ input_per_million: {
1292
+ threshold_tokens: 200000,
1293
+ price_below_threshold_per_million: 3.0,
1294
+ price_above_threshold_per_million: 6.0,
1295
+ },
1296
+ output_per_million: {
1297
+ threshold_tokens: 200000,
1298
+ price_below_threshold_per_million: 15.0,
1299
+ price_above_threshold_per_million: 22.5,
1300
+ },
1301
+ cached_input_per_million: {
1302
+ threshold_tokens: 200000,
1303
+ price_below_threshold_per_million: 0.3,
1304
+ price_above_threshold_per_million: 0.6,
1305
+ },
1306
+ },
1307
+ features: {
1308
+ context_length: 200000,
1309
+ input_modality: ['text', 'image'],
1310
+ output_modality: ['text'],
1311
+ tool_use: true,
1312
+ streaming: true,
1313
+ json_output: true,
1314
+ reasoning_output: true,
1315
+ max_output_tokens: 64000,
1316
+ },
1317
+ class: 'reasoning',
1318
+ score: 88,
1319
+ description: 'Claude Sonnet 4.5 (Sep 2025 refresh) with tiered pricing at 200k tokens and vision support.',
1320
+ },
1321
+ {
1322
+ id: 'claude-sonnet-4-5-20250514-long-context',
1323
+ aliases: ['claude-sonnet-4-5-long', 'claude-sonnet-4.5-long'],
1324
+ provider: 'anthropic',
1325
+ cost: {
1326
+ input_per_million: 6.0,
1327
+ output_per_million: 22.5,
1328
+ cached_input_per_million: 0.6,
1329
+ },
1330
+ features: {
1331
+ context_length: 1_000_000,
1332
+ input_modality: ['text', 'image'],
1333
+ output_modality: ['text'],
1334
+ tool_use: true,
1335
+ streaming: true,
1336
+ json_output: true,
1337
+ reasoning_output: true,
1338
+ max_output_tokens: 64000,
1339
+ },
1340
+ class: 'reasoning',
1341
+ score: 88,
1342
+ description: 'Claude Sonnet 4.5 with 1M token context window - for long-context processing',
1343
+ },
1344
+ {
1345
+ id: 'claude-haiku-4-5-20250514',
1346
+ aliases: ['claude-haiku-4-5', 'claude-haiku-4.5', 'claude-haiku-latest'],
1347
+ provider: 'anthropic',
1348
+ cost: {
1349
+ input_per_million: 1.0,
1350
+ output_per_million: 5.0,
1351
+ cached_input_per_million: 0.1,
1024
1352
  },
1025
1353
  features: {
1026
1354
  context_length: 200000,
@@ -1032,21 +1360,22 @@ exports.MODEL_REGISTRY = [
1032
1360
  max_output_tokens: 8192,
1033
1361
  },
1034
1362
  class: 'mini',
1035
- score: 70,
1363
+ score: 74,
1036
1364
  scores: {
1037
- monologue: 66,
1038
- code: 63,
1039
- reasoning: 55,
1365
+ monologue: 72,
1366
+ code: 71,
1367
+ reasoning: 70,
1040
1368
  },
1041
- description: 'Fast, cost-effective Claude model',
1369
+ description: 'Claude Haiku 4.5 - Fast, cost-effective model with text and image support',
1042
1370
  },
1043
1371
  {
1044
- id: 'claude-cli',
1372
+ id: 'claude-haiku-4-5-20251001',
1373
+ aliases: ['claude-haiku-4-5-oct-2025', 'claude-haiku-4.5-2025-10-01'],
1045
1374
  provider: 'anthropic',
1046
1375
  cost: {
1047
- input_per_million: 3.0,
1048
- output_per_million: 15.0,
1049
- cached_input_per_million: 0.3,
1376
+ input_per_million: 1.0,
1377
+ output_per_million: 5.0,
1378
+ cached_input_per_million: 0.1,
1050
1379
  },
1051
1380
  features: {
1052
1381
  context_length: 200000,
@@ -1055,18 +1384,19 @@ exports.MODEL_REGISTRY = [
1055
1384
  tool_use: true,
1056
1385
  streaming: true,
1057
1386
  json_output: true,
1387
+ max_output_tokens: 8192,
1058
1388
  },
1059
- class: 'reasoning',
1060
- description: 'Claude accessed via CLI (likely uses latest Sonnet or Haiku model)',
1389
+ class: 'mini',
1390
+ score: 74,
1391
+ description: 'Claude Haiku 4.5 (Oct 2025 refresh) fast tier with vision support.',
1061
1392
  },
1062
1393
  {
1063
- id: 'claude-opus-4-1-20250805',
1064
- aliases: ['claude-opus-4', 'claude-opus-4-1', 'claude-4-opus', 'claude-opus-4-20250514'],
1394
+ id: 'claude-cli',
1065
1395
  provider: 'anthropic',
1066
1396
  cost: {
1067
- input_per_million: 15.0,
1068
- output_per_million: 75.0,
1069
- cached_input_per_million: 1.5,
1397
+ input_per_million: 3.0,
1398
+ output_per_million: 15.0,
1399
+ cached_input_per_million: 0.3,
1070
1400
  },
1071
1401
  features: {
1072
1402
  context_length: 200000,
@@ -1075,21 +1405,18 @@ exports.MODEL_REGISTRY = [
1075
1405
  tool_use: true,
1076
1406
  streaming: true,
1077
1407
  json_output: true,
1078
- max_output_tokens: 32000,
1079
- reasoning_output: true,
1080
1408
  },
1081
1409
  class: 'reasoning',
1082
- score: 95,
1083
- description: 'Claude Opus 4 - Highest level of intelligence and capability with extended thinking',
1410
+ description: 'Claude accessed via CLI (likely uses latest Sonnet or Haiku model)',
1084
1411
  },
1085
1412
  {
1086
- id: 'claude-sonnet-4-20250514',
1087
- aliases: ['claude-sonnet-4', 'claude-4-sonnet'],
1413
+ id: 'claude-opus-4-1-20250805',
1414
+ aliases: ['claude-opus-4-1', 'claude-opus-4.1', 'claude-4-opus', 'claude-opus'],
1088
1415
  provider: 'anthropic',
1089
1416
  cost: {
1090
- input_per_million: 3.0,
1091
- output_per_million: 15.0,
1092
- cached_input_per_million: 0.3,
1417
+ input_per_million: 15.0,
1418
+ output_per_million: 75.0,
1419
+ cached_input_per_million: 1.5,
1093
1420
  },
1094
1421
  features: {
1095
1422
  context_length: 200000,
@@ -1098,28 +1425,26 @@ exports.MODEL_REGISTRY = [
1098
1425
  tool_use: true,
1099
1426
  streaming: true,
1100
1427
  json_output: true,
1101
- max_output_tokens: 64000,
1102
1428
  reasoning_output: true,
1429
+ max_output_tokens: 32000,
1103
1430
  },
1104
1431
  class: 'reasoning',
1105
- score: 90,
1106
- description: 'Claude Sonnet 4 - High intelligence and balanced performance with extended thinking',
1432
+ score: 95,
1433
+ scores: {
1434
+ monologue: 94,
1435
+ code: 94,
1436
+ reasoning: 93,
1437
+ },
1438
+ description: 'Claude Opus 4.1 - Highest intelligence and capability with reasoning support',
1107
1439
  },
1108
1440
  {
1109
1441
  id: 'gemini-2.5-pro-preview-06-05',
1110
1442
  aliases: ['gemini-2.5-pro', 'gemini-2.5-pro-exp-03-25', 'gemini-2.5-pro-preview-05-06'],
1111
1443
  provider: 'google',
1112
1444
  cost: {
1113
- input_per_million: {
1114
- threshold_tokens: 200000,
1115
- price_below_threshold_per_million: 1.25,
1116
- price_above_threshold_per_million: 2.5,
1117
- },
1118
- output_per_million: {
1119
- threshold_tokens: 200000,
1120
- price_below_threshold_per_million: 10.0,
1121
- price_above_threshold_per_million: 15.0,
1122
- },
1445
+ input_per_million: 1.25,
1446
+ output_per_million: 10.0,
1447
+ cached_input_per_million: 0.13,
1123
1448
  },
1124
1449
  features: {
1125
1450
  context_length: 1048576,
@@ -1139,8 +1464,11 @@ exports.MODEL_REGISTRY = [
1139
1464
  aliases: ['gemini-2.5-flash', 'gemini-2.5-flash-preview-04-17'],
1140
1465
  provider: 'google',
1141
1466
  cost: {
1142
- input_per_million: 0.3,
1143
- output_per_million: 2.5,
1467
+ input_per_million: {
1468
+ text: 1.0,
1469
+ audio: 0.0375,
1470
+ },
1471
+ output_per_million: 0.6,
1144
1472
  },
1145
1473
  features: {
1146
1474
  context_length: 1048576,
@@ -1165,8 +1493,8 @@ exports.MODEL_REGISTRY = [
1165
1493
  aliases: ['gemini-2.5-flash-lite'],
1166
1494
  provider: 'google',
1167
1495
  cost: {
1168
- input_per_million: 0.1,
1169
- output_per_million: 0.4,
1496
+ input_per_million: 0.05,
1497
+ output_per_million: 0.2,
1170
1498
  },
1171
1499
  features: {
1172
1500
  context_length: 1000000,
@@ -1178,13 +1506,59 @@ exports.MODEL_REGISTRY = [
1178
1506
  max_output_tokens: 64000,
1179
1507
  },
1180
1508
  class: 'reasoning_mini',
1181
- score: 75,
1509
+ score: 78,
1182
1510
  scores: {
1183
- monologue: 12,
1184
- code: 63,
1185
- reasoning: 78,
1511
+ monologue: 14,
1512
+ code: 65,
1513
+ reasoning: 75,
1186
1514
  },
1187
- description: 'Balanced multimodal model with large context, built for Agents.',
1515
+ description: 'Gemini 2.5 Flash Lite - cost-efficient multimodal model with 1M context.',
1516
+ },
1517
+ {
1518
+ id: 'gemini-3-pro-preview',
1519
+ aliases: ['gemini-3-pro', 'gemini-3-pro-preview-11-2025'],
1520
+ provider: 'google',
1521
+ cost: {
1522
+ input_per_million: {
1523
+ threshold_tokens: 200000,
1524
+ price_below_threshold_per_million: 2.0,
1525
+ price_above_threshold_per_million: 4.0,
1526
+ },
1527
+ output_per_million: {
1528
+ threshold_tokens: 200000,
1529
+ price_below_threshold_per_million: 12.0,
1530
+ price_above_threshold_per_million: 18.0,
1531
+ },
1532
+ },
1533
+ features: {
1534
+ context_length: 1_000_000,
1535
+ input_modality: ['text', 'image', 'video', 'audio'],
1536
+ output_modality: ['text', 'image', 'audio'],
1537
+ tool_use: true,
1538
+ streaming: true,
1539
+ json_output: true,
1540
+ max_output_tokens: 8192,
1541
+ },
1542
+ class: 'reasoning',
1543
+ score: 90,
1544
+ description: 'Gemini 3 Pro Preview - multimodal flagship (v1beta, tiered pricing at 200k tokens).',
1545
+ },
1546
+ {
1547
+ id: 'gemini-3-pro-image-preview',
1548
+ aliases: ['gemini-3-pro-image'],
1549
+ provider: 'google',
1550
+ cost: {
1551
+ per_image: 0.134,
1552
+ input_per_million: 2.0,
1553
+ output_per_million: 12.0,
1554
+ },
1555
+ features: {
1556
+ input_modality: ['text'],
1557
+ output_modality: ['image', 'text'],
1558
+ streaming: false,
1559
+ },
1560
+ class: 'image_generation',
1561
+ description: 'Gemini 3 Pro Image (preview) for text-to-image generation.',
1188
1562
  },
1189
1563
  {
1190
1564
  id: 'gemini-2.0-flash-lite',
@@ -1373,6 +1747,22 @@ exports.MODEL_REGISTRY = [
1373
1747
  class: 'voice',
1374
1748
  description: "Gemini's advanced text-to-speech model with superior voice quality, expression control, and multi-speaker support for creating dynamic conversations.",
1375
1749
  },
1750
+ {
1751
+ id: 'gemini-2.5-flash-native-audio-preview',
1752
+ provider: 'google',
1753
+ cost: {
1754
+ input_per_million: 5.0,
1755
+ output_per_million: 0,
1756
+ },
1757
+ features: {
1758
+ input_modality: ['text'],
1759
+ output_modality: ['audio'],
1760
+ streaming: true,
1761
+ context_length: 32000,
1762
+ },
1763
+ class: 'voice',
1764
+ description: "Gemini's native audio preview model providing low-latency text-to-speech with built-in audio processing.",
1765
+ },
1376
1766
  {
1377
1767
  id: 'codex-mini-latest',
1378
1768
  provider: 'openai',