diff --git a/litellm/integrations/otel/emitter.py b/litellm/integrations/otel/emitter.py
index 1a751973eac..d2cae2766a9 100644
--- a/litellm/integrations/otel/emitter.py
+++ b/litellm/integrations/otel/emitter.py
@@ -7,6 +7,7 @@ from typing import Final
 
 from opentelemetry.context import Context
 from opentelemetry.sdk.trace import ReadableSpan, SpanLimits
+from opentelemetry.sdk.trace import Tracer as SdkTracer
 from opentelemetry.trace import Link, Span, Tracer
 from opentelemetry.trace.status import Status, StatusCode
 
@@ -83,6 +84,13 @@ def error_attributes(error: SpanError) -> Mapping[str, AttrValue]:
     return MappingProxyType({key: value for key, value in pairs if value})
 
 
+def span_attribute_limit(tracer: Tracer) -> int | None:
+    """The attribute count limit spans started by ``tracer`` are built with, ``None`` when unbounded."""
+    if not isinstance(tracer, SdkTracer):
+        return SpanLimits().max_span_attributes
+    return tracer._span_limits.max_span_attributes  # pyright: ignore[reportPrivateUsage]  # SDK has no public getter
+
+
 def stamp_error(
     span: Span,
     error: SpanError,
@@ -126,14 +134,11 @@ class SpanEmitter:
         config: OpenTelemetryV2Config,
         mappers: Sequence[AttributeMapper] | None = None,
         event_recorder: GenAIEventRecorder | None = None,
-        span_attribute_limit: int | None = None,
     ) -> None:
         self._tracer = tracer
         self._config = config
         self._event_recorder = event_recorder
-        self._span_attribute_limit: int | None = (
-            SpanLimits().max_span_attributes if span_attribute_limit is None else span_attribute_limit
-        )
+        self._span_attribute_limit: int | None = span_attribute_limit(tracer)
         # The mapper chain is the sole source of span attributes. When not
         # passed in, resolve it from the config so there's one source of truth.
         self._mappers: list[AttributeMapper] = (
@@ -269,7 +274,8 @@ class SpanEmitter:
         mapped: Final = MappingProxyType(
             {key: value for mapper in self._mappers for key, value in mapper.map(data).items()}
         )
-        reserved: Final = len(error_attributes(error)) if error else 0
+        stamped_later: Final = error_attributes(error) if error else _NO_ATTRIBUTES
+        reserved: Final = len(stamped_later.keys() - mapped.keys())
         for key, value in fit_indexed_messages(mapped, self._attribute_budget(span, reserved)).items():
             span.set_attribute(key, value)
         if error:
diff --git a/tests/test_litellm/integrations/otel/test_otel_v2_emitter.py b/tests/test_litellm/integrations/otel/test_otel_v2_emitter.py
index 74b031f7f09..828759d2f38 100644
--- a/tests/test_litellm/integrations/otel/test_otel_v2_emitter.py
+++ b/tests/test_litellm/integrations/otel/test_otel_v2_emitter.py
@@ -7,8 +7,10 @@ import pytest
 
 pytest.importorskip("opentelemetry")
 
-from opentelemetry.sdk.trace import SpanLimits  # noqa: E402
-from opentelemetry.trace import SpanKind  # noqa: E402
+from opentelemetry.sdk.trace import SpanLimits, TracerProvider  # noqa: E402
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor  # noqa: E402
+from opentelemetry.sdk.trace.export.in_memory_span_exporter import InMemorySpanExporter  # noqa: E402
+from opentelemetry.trace import NoOpTracer, SpanKind  # noqa: E402
 from opentelemetry.trace.status import StatusCode  # noqa: E402
 
 from litellm.integrations.otel import (  # noqa: E402
@@ -18,7 +20,7 @@ from litellm.integrations.otel import (  # noqa: E402
 )
 from litellm.integrations.otel.plumbing import context as ctx_mod  # noqa: E402
 from litellm.integrations.otel.plumbing import providers  # noqa: E402
-from litellm.integrations.otel.emitter import SpanEmitter  # noqa: E402
+from litellm.integrations.otel.emitter import SpanEmitter, span_attribute_limit  # noqa: E402
 from litellm.integrations.otel.emitter import stamp_error  # noqa: E402
 from litellm.integrations.otel.mappers.utils import MAX_TOOL_DEFINITION_ATTRS_PER_SPAN  # noqa: E402
 from litellm.integrations.otel.model.payloads import (  # noqa: E402
@@ -439,15 +441,19 @@ def _conversation_payload(turns, choices=1, **overrides):
     )
 
 
-def _conversation_span(mapper_names, payload, legacy_compat=False):
-    """The exported LLM-call span for ``payload`` with content capture on."""
+def _conversation_span(mapper_names, payload, legacy_compat=False, span_limits=None):
+    """The exported LLM-call span for ``payload`` with content capture on.
+
+    ``span_limits`` builds the provider with programmatic limits instead of the environment's."""
     cfg = OpenTelemetryV2Config(
         exporter="in_memory",
         legacy_compat=legacy_compat,
         mapper_names=list(mapper_names),
         capture_message_content="span_only",
     )
-    provider, exporter = providers.in_memory_provider(cfg)
+    provider, exporter = (
+        providers.in_memory_provider(cfg) if span_limits is None else _provider_with_limits(span_limits)
+    )
     engine = SpanEmitter(providers.get_tracer(provider, "litellm-test"), cfg)
     engine.emit(
         SpanRole.LLM_CALL,
@@ -457,6 +463,13 @@ def _conversation_span(mapper_names, payload, legacy_compat=False):
     return span
 
 
+def _provider_with_limits(span_limits):
+    provider = TracerProvider(span_limits=span_limits)
+    exporter = InMemorySpanExporter()
+    provider.add_span_processor(SimpleSpanProcessor(exporter))
+    return provider, exporter
+
+
 def _indexed_messages(attributes, prefix):
     return sorted({int(key.split(".")[2]) for key in attributes if key.startswith(f"{prefix}.")})
 
@@ -617,7 +630,7 @@ def test_error_and_pre_stamped_attributes_keep_their_room_on_a_long_conversation
     a = s.attributes
 
     assert s.dropped_attributes == 0
-    assert len(a) <= SpanLimits().max_span_attributes
+    assert SpanLimits().max_span_attributes - 1 <= len(a) <= SpanLimits().max_span_attributes
     assert a[GenAI.REQUEST_MODEL] == "gpt-4o"
     assert a["litellm.metadata.baggage_0"] == "value 0"
     assert a["error.type"] == "RateLimitError"
@@ -626,6 +639,34 @@ def test_error_and_pre_stamped_attributes_keep_their_room_on_a_long_conversation
     assert a["llm.input_messages.59.message.content"] == "turn 59"
 
 
+def test_indexed_messages_follow_the_providers_own_span_limits(monkeypatch):
+    """A provider built with programmatic ``SpanLimits`` sets the budget, whatever the environment says."""
+    monkeypatch.setenv("OTEL_SPAN_ATTRIBUTE_COUNT_LIMIT", "1000")
+    span = _conversation_span(
+        ["genai", "openinference"], _conversation_payload(60), span_limits=SpanLimits(max_span_attributes=40)
+    )
+    _assert_core_intact(span)
+    a = span.attributes
+    assert 39 <= len(a) <= 40
+    assert a["llm.input_messages.0.message.content"] == "turn 0"
+    assert a["llm.input_messages.59.message.content"] == "turn 59"
+    assert a["llm.output_messages.0.message.content"] == "reply 0"
+
+    monkeypatch.setenv("OTEL_SPAN_ATTRIBUTE_COUNT_LIMIT", "48")
+    unbounded = _conversation_span(
+        ["genai", "openinference"],
+        _conversation_payload(60),
+        span_limits=SpanLimits(max_span_attributes=SpanLimits.UNSET),
+    )
+    _assert_core_intact(unbounded)
+    assert _indexed_messages(unbounded.attributes, "llm.input_messages") == list(range(60))
+
+
+def test_span_attribute_limit_falls_back_to_the_environment_for_tracers_outside_the_sdk(monkeypatch):
+    monkeypatch.setenv("OTEL_SPAN_ATTRIBUTE_COUNT_LIMIT", "48")
+    assert span_attribute_limit(NoOpTracer()) == 48
+
+
 def test_fully_populated_span_with_every_vocabulary_stays_within_the_attribute_limit():
     """Every capped family maxed at once still leaves the whole core intact."""
     payload = _conversation_payload(