commit e56550c435b8bd6351241b5c902eec03edc5f344
parent 21780745c11f773cbe8f7046a012762ae0fc812e
Author: andrewlaack-collab <andrew.laack@imbue.com>
Date: Wed, 25 Feb 2026 16:24:27 -0600
Anthropic model updates (#147)
Co-authored-by: Andrew Laack <andrew@laack.co>
Co-authored-by: OpenCode <opencode@users.noreply.github.com>
Diffstat:
4 files changed, 43 insertions(+), 139 deletions(-)
diff --git a/.vet/models.json b/.vet/models.json
@@ -45,7 +45,7 @@
"supports_temperature": true
},
"haiku": {
- "model_id": "claude-haiku-4-5-20251001",
+ "model_id": "claude-haiku-4-5",
"context_window": 200000,
"max_output_tokens": 16384,
"supports_temperature": true
diff --git a/vet/imbue_core/agents/llm_apis/anthropic_api.py b/vet/imbue_core/agents/llm_apis/anthropic_api.py
@@ -55,32 +55,21 @@ from vet.imbue_core.secrets_utils import get_secret
class AnthropicModelName(enum.StrEnum):
- CLAUDE_3_HAIKU_2024_03_07 = "claude-3-haiku-20240307"
- CLAUDE_3_OPUS_2024_02_29 = "claude-3-opus-20240229"
- CLAUDE_3_5_SONNET_2024_06_20 = "claude-3-5-sonnet-20240620"
- CLAUDE_3_5_SONNET_2024_10_22 = "claude-3-5-sonnet-20241022"
- CLAUDE_3_5_HAIKU_2024_10_22 = "claude-3-5-haiku-20241022"
- CLAUDE_3_7_SONNET_2025_02_19 = "claude-3-7-sonnet-20250219"
- CLAUDE_4_OPUS_2025_05_14 = "claude-opus-4-20250514"
- CLAUDE_4_1_OPUS_2025_08_05 = "claude-opus-4-1-20250805"
- CLAUDE_4_SONNET_2025_05_14 = "claude-sonnet-4-20250514"
- CLAUDE_4_5_SONNET_2025_09_29 = "claude-sonnet-4-5-20250929"
- CLAUDE_4_5_HAIKU_2025_10_01 = "claude-haiku-4-5-20251001"
- CLAUDE_4_5_OPUS_2025_11_01 = "claude-opus-4-5-20251101"
+ CLAUDE_4_OPUS = "claude-opus-4-0"
+ CLAUDE_4_1_OPUS = "claude-opus-4-1"
+ CLAUDE_4_5_OPUS = "claude-opus-4-5"
CLAUDE_4_6_OPUS = "claude-opus-4-6"
+ CLAUDE_4_SONNET = "claude-sonnet-4-0"
+ CLAUDE_4_5_SONNET = "claude-sonnet-4-5"
CLAUDE_4_6_SONNET = "claude-sonnet-4-6"
+ CLAUDE_4_5_HAIKU = "claude-haiku-4-5"
# the same as above but with the token limit and cost per token for the 1M token limit
# TODO: combine these and add ability for token costs to be nonlinear
# FIXME: this is an exception where the model name is not the same as the model name in the API
- CLAUDE_4_SONNET_2025_05_14_LONG = "claude-sonnet-4-20250514-long"
- CLAUDE_4_5_SONNET_2025_09_29_LONG = "claude-sonnet-4-5-20250929-long"
+ CLAUDE_4_SONNET_LONG = "claude-sonnet-4-0-long"
+ CLAUDE_4_5_SONNET_LONG = "claude-sonnet-4-5-long"
CLAUDE_4_6_OPUS_LONG = "claude-opus-4-6-long"
- # the following are 'retired' and are no longer available: https://docs.claude.com/en/docs/about-claude/model-deprecations
- # CLAUDE_2_1 = "claude-2.1"
- # CLAUDE_2 = "claude-2"
- # CLAUDE_3_SONNET_2024_02_29 = "claude-3-sonnet-20240229"
-
# Basic info is available at https://docs.anthropic.com/claude/reference/models
# Rate limits for Anthropic models are available on our dashboard: https://console.anthropic.com/settings/limits
@@ -89,93 +78,8 @@ class AnthropicModelName(enum.StrEnum):
# NOTE: as of 2025-06-04, there are some models that don't have rate limits set in our dashboard
ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = FrozenDict(
{
- AnthropicModelName.CLAUDE_3_HAIKU_2024_03_07: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_HAIKU_2024_03_07,
- cost_per_input_token=0.25 / 1_000_000,
- cost_per_output_token=1.25 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=4096,
- rate_limit_req=4000 / 60, # 4000 RPM = 66.67 RPS
- rate_limit_tok=4_000_000 / 60,
- rate_limit_output_tok=800_000 / 60,
- provider_specific_info=AnthropicModelInfo(
- cost_per_5m_cache_write_token=0.3 / 1_000_000,
- cost_per_1h_cache_write_token=0.5 / 1_000_000,
- cost_per_cache_read_token=0.03 / 1_000_000,
- ),
- ),
- AnthropicModelName.CLAUDE_3_OPUS_2024_02_29: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_OPUS_2024_02_29,
- cost_per_input_token=15.00 / 1_000_000,
- cost_per_output_token=75.00 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=4096,
- rate_limit_req=4000 / 60, # 4000 RPM = 66.67 RPS
- rate_limit_tok=1_000_000 / 60,
- rate_limit_output_tok=150_000 / 60,
- provider_specific_info=AnthropicModelInfo(
- cost_per_5m_cache_write_token=18.75 / 1_000_000,
- cost_per_1h_cache_write_token=30 / 1_000_000,
- cost_per_cache_read_token=1.5 / 1_000_000,
- ),
- ),
- AnthropicModelName.CLAUDE_3_5_SONNET_2024_06_20: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_5_SONNET_2024_06_20,
- cost_per_input_token=3.00 / 1_000_000,
- cost_per_output_token=15.00 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=4096,
- rate_limit_req=5000 / 60, # 5000 RPM = 83.33 RPS
- rate_limit_tok=8_000_000 / 60,
- rate_limit_output_tok=1_600_000 / 60,
- provider_specific_info=AnthropicModelInfo(
- cost_per_5m_cache_write_token=3.75 / 1_000_000,
- cost_per_1h_cache_write_token=6 / 1_000_000,
- cost_per_cache_read_token=0.3 / 1_000_000,
- ),
- ),
- AnthropicModelName.CLAUDE_3_5_SONNET_2024_10_22: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_5_SONNET_2024_10_22,
- cost_per_input_token=3.00 / 1_000_000,
- cost_per_output_token=15.00 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=8192,
- rate_limit_req=5000 / 60, # 5000 RPM = 83.33 RPS
- rate_limit_tok=8_000_000 / 60,
- rate_limit_output_tok=400_000 / 60,
- ),
- AnthropicModelName.CLAUDE_3_5_HAIKU_2024_10_22: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_5_HAIKU_2024_10_22,
- cost_per_input_token=1.00 / 1_000_000,
- cost_per_output_token=5.00 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=8192,
- rate_limit_req=4000 / 60, # 4000 RPM = 66.67 RPS
- rate_limit_tok=4_000_000 / 60,
- rate_limit_output_tok=800_000 / 60,
- provider_specific_info=AnthropicModelInfo(
- cost_per_5m_cache_write_token=1 / 1_000_000,
- cost_per_1h_cache_write_token=1.6 / 1_000_000,
- cost_per_cache_read_token=0.08 / 1_000_000,
- ),
- ),
- AnthropicModelName.CLAUDE_3_7_SONNET_2025_02_19: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_3_7_SONNET_2025_02_19,
- cost_per_input_token=3.00 / 1_000_000,
- cost_per_output_token=15.00 / 1_000_000,
- max_input_tokens=200_000,
- max_output_tokens=8192,
- rate_limit_req=None, # Currently no limit set in our dashboard
- rate_limit_tok=2_000_000 / 60,
- rate_limit_output_tok=400_000 / 60,
- provider_specific_info=AnthropicModelInfo(
- cost_per_5m_cache_write_token=3.75 / 1_000_000,
- cost_per_1h_cache_write_token=6 / 1_000_000,
- cost_per_cache_read_token=0.3 / 1_000_000,
- ),
- ),
- AnthropicModelName.CLAUDE_4_OPUS_2025_05_14: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_OPUS_2025_05_14,
+ AnthropicModelName.CLAUDE_4_OPUS: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_OPUS,
cost_per_input_token=15.00 / 1_000_000,
cost_per_output_token=75.00 / 1_000_000,
max_input_tokens=200_000,
@@ -189,8 +93,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=1.5 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_1_OPUS_2025_08_05: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_1_OPUS_2025_08_05,
+ AnthropicModelName.CLAUDE_4_1_OPUS: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_1_OPUS,
cost_per_input_token=15.00 / 1_000_000,
cost_per_output_token=75.00 / 1_000_000,
max_input_tokens=200_000,
@@ -204,8 +108,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=1.5 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_5_OPUS_2025_11_01: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_5_OPUS_2025_11_01,
+ AnthropicModelName.CLAUDE_4_5_OPUS: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_5_OPUS,
cost_per_input_token=5.00 / 1_000_000,
cost_per_output_token=25.00 / 1_000_000,
max_input_tokens=200_000,
@@ -234,8 +138,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=0.50 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_SONNET_2025_05_14: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_SONNET_2025_05_14,
+ AnthropicModelName.CLAUDE_4_SONNET: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_SONNET,
cost_per_input_token=3.00 / 1_000_000,
cost_per_output_token=15.00 / 1_000_000,
max_input_tokens=200_000,
@@ -249,8 +153,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=0.3 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29,
+ AnthropicModelName.CLAUDE_4_5_SONNET: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_5_SONNET,
cost_per_input_token=3.00 / 1_000_000,
cost_per_output_token=15.00 / 1_000_000,
max_input_tokens=200_000,
@@ -279,8 +183,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=0.3 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_5_HAIKU_2025_10_01: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_5_HAIKU_2025_10_01,
+ AnthropicModelName.CLAUDE_4_5_HAIKU: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_5_HAIKU,
cost_per_input_token=1.00 / 1_000_000,
cost_per_output_token=5.00 / 1_000_000,
max_input_tokens=200_000,
@@ -294,8 +198,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
cost_per_cache_read_token=0.1 / 1_000_000,
),
),
- AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG,
+ AnthropicModelName.CLAUDE_4_SONNET_LONG: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_SONNET_LONG,
# the first 200_000 input tokens use the rates above, and the next up to 800_000 use the rate 6.0 / 1_000_000.
# thus the maximum average cost per input token is (3.0 * 200_000 + 6.0 * 800_000) / 1_000_000 = 5.4 per 1_000_000.
# (all output tokens may be past 200_000 input tokens, so the max average cost there is just the cost for tokens after 200_000)
@@ -307,8 +211,8 @@ ANTHROPIC_MODEL_INFO_BY_NAME: FrozenMapping[AnthropicModelName, ModelInfo] = Fro
rate_limit_tok=1_000_000 / 60, # <-- yeah they let us have one (1) 1M request per minute
rate_limit_output_tok=200_000 / 60,
),
- AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG: ModelInfo(
- model_name=AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG,
+ AnthropicModelName.CLAUDE_4_5_SONNET_LONG: ModelInfo(
+ model_name=AnthropicModelName.CLAUDE_4_5_SONNET_LONG,
# the first 200_000 input tokens use the rates above, and the next up to 800_000 use the rate 6.0 / 1_000_000.
# thus the maximum average cost per input token is (3.0 * 200_000 + 6.0 * 800_000) / 1_000_000 = 5.4 per 1_000_000.
# (all output tokens may be past 200_000 input tokens, so the max average cost there is just the cost for tokens after 200_000)
@@ -484,7 +388,7 @@ class MissingCachingInfoError(Exception):
class AnthropicAPI(LanguageModelAPI):
- model_name: AnthropicModelName = AnthropicModelName.CLAUDE_4_SONNET_2025_05_14
+ model_name: AnthropicModelName = AnthropicModelName.CLAUDE_4_SONNET
is_conversational: bool = True
# Anthropic specific args
@@ -559,16 +463,16 @@ class AnthropicAPI(LanguageModelAPI):
assert params.max_tokens is not None, "max_tokens must be provided for Anthropic API"
if self.model_name in (
- AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG,
- AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG,
+ AnthropicModelName.CLAUDE_4_5_SONNET_LONG,
+ AnthropicModelName.CLAUDE_4_SONNET_LONG,
AnthropicModelName.CLAUDE_4_6_OPUS_LONG,
):
# FIXME: Fix this once this is no longer beta or as this becomes required for more models
# Map the name back to the actual model name for the API call
- if self.model_name == AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG:
- model_name = AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29
- elif self.model_name == AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG:
- model_name = AnthropicModelName.CLAUDE_4_SONNET_2025_05_14
+ if self.model_name == AnthropicModelName.CLAUDE_4_5_SONNET_LONG:
+ model_name = AnthropicModelName.CLAUDE_4_5_SONNET
+ elif self.model_name == AnthropicModelName.CLAUDE_4_SONNET_LONG:
+ model_name = AnthropicModelName.CLAUDE_4_SONNET
elif self.model_name == AnthropicModelName.CLAUDE_4_6_OPUS_LONG:
model_name = AnthropicModelName.CLAUDE_4_6_OPUS
else:
@@ -644,16 +548,16 @@ class AnthropicAPI(LanguageModelAPI):
assert max_tokens is not None, "max_tokens must be provided for Anthropic API"
if self.model_name in (
- AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG,
- AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG,
+ AnthropicModelName.CLAUDE_4_5_SONNET_LONG,
+ AnthropicModelName.CLAUDE_4_SONNET_LONG,
AnthropicModelName.CLAUDE_4_6_OPUS_LONG,
):
# FIXME: Fix this once this is no longer beta or as this becomes required for more models
# Map the name back to the actual model name for the API call
- if self.model_name == AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29_LONG:
- model_name = AnthropicModelName.CLAUDE_4_5_SONNET_2025_09_29
- elif self.model_name == AnthropicModelName.CLAUDE_4_SONNET_2025_05_14_LONG:
- model_name = AnthropicModelName.CLAUDE_4_SONNET_2025_05_14
+ if self.model_name == AnthropicModelName.CLAUDE_4_5_SONNET_LONG:
+ model_name = AnthropicModelName.CLAUDE_4_5_SONNET
+ elif self.model_name == AnthropicModelName.CLAUDE_4_SONNET_LONG:
+ model_name = AnthropicModelName.CLAUDE_4_SONNET
elif self.model_name == AnthropicModelName.CLAUDE_4_6_OPUS_LONG:
model_name = AnthropicModelName.CLAUDE_4_6_OPUS
else:
diff --git a/vet/imbue_core/agents/llm_apis/common.py b/vet/imbue_core/agents/llm_apis/common.py
@@ -46,12 +46,12 @@ def get_all_model_names() -> list[str]:
def get_formatted_model_name(model_name: str) -> str:
"""Get a nicely formatted model name.
- Does things like removing generic prefixes like 'models/' and forward slashes (which can interfere with file names).
+ Does things like removing generic prefixes like 'models/' and forward slashes (which can interfere with file names).
- Some examples:
+ Some examples:
- - `models/gemini-1.5-flash-001` -> `gemini-1.5-flash-001`
- - 'claude-3-5-haiku-20241022' -> 'claude-3-5-haiku-20241022'
+ - `models/gemini-1.5-flash-001` -> `gemini-1.5-flash-001`
+ - 'claude-opus-4-6' -> 'claude-opus-4-6'
"""
if model_name.startswith("models/"):
diff --git a/vet/repo_utils_test.py b/vet/repo_utils_test.py
@@ -81,7 +81,7 @@ def test_build_context(simple_test_git_repo: Path, snapshot: SnapshotAssertion)
project_context = LazyProjectContext.build(
git_hash,
diff,
- language_model_name=AnthropicModelName.CLAUDE_4_5_HAIKU_2025_10_01,
+ language_model_name=AnthropicModelName.CLAUDE_4_5_HAIKU,
repo_path=simple_test_git_repo,
tokens_to_reserve=20000,
).to_base_project_context()
